BEGIN:VCALENDAR VERSION:2.0 PRODID:-//HasGeek//NONSGML Funnel//EN DESCRIPTION:A conference on data\, machine learning\, and distributed and parallel computing X-WR-CALDESC:A conference on data\, machine learning\, and distributed and parallel computing NAME:The Fifth Elephant 2015 X-WR-CALNAME:The Fifth Elephant 2015 REFRESH-INTERVAL;VALUE=DURATION:PT12H SUMMARY:The Fifth Elephant 2015 TIMEZONE-ID:Asia/Kolkata X-PUBLISHED-TTL:PT12H X-WR-TIMEZONE:Asia/Kolkata BEGIN:VEVENT SUMMARY:Check-in DTSTART:20150716T030000Z DTEND:20150716T035500Z DTSTAMP:20260423T114800Z UID:session/WTpzuiwjcvG1PFYAx14xmc@hasgeek.com SEQUENCE:0 CREATED:20150522T141729Z DESCRIPTION:\n LAST-MODIFIED:20150705T131826Z LOCATION:Bangalore ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com BEGIN:VALARM ACTION:display DESCRIPTION:Check-in in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Conference Introduction DTSTART:20150716T035500Z DTEND:20150716T041500Z DTSTAMP:20260423T114800Z UID:session/T15JEeo5JzX3wJEDMw95jf@hasgeek.com SEQUENCE:0 CREATED:20150522T141746Z DESCRIPTION:\n GEO:12.943163506799058;77.596340412274 LAST-MODIFIED:20150705T131818Z LOCATION:Audi 1 - NIMHANS Convention center\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com BEGIN:VALARM ACTION:display DESCRIPTION:Conference Introduction in Audi 1 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Graph Algorithms and Computer Vision DTSTART:20150716T041500Z DTEND:20150716T050000Z DTSTAMP:20260423T114800Z UID:session/9odPE4r4tFV5wENa3iER7d@hasgeek.com SEQUENCE:2 CATEGORIES:Full Talk,Intermediate CREATED:20150705T132659Z DESCRIPTION:Graph based algorithms have found extensive applications in Co mputer Vision and have been in forefront for many vision applications. The y have shown to have connections to other areas such as Natural Language P rocessing and Network Analysis. This talk aims to showcase some of the kno wn connections between these areas with applications to data\, big and sma ll. This will be centered around applications in Computer Vision such as s egmentation and face detection but will make forays into ranking and clust ering. This would be a survey talk and will be aimed at showcasing the int uition behind some of these algorithms as opposed to enunciating the rigor ous mathematics.\n\n### Speaker bio\n\nhttps://in.linkedin.com/pub/sumod-k -mohan/7/12/3a3\n GEO:12.943163506799058;77.596340412274 LAST-MODIFIED:20230810T072606Z LOCATION:Audi 2 - NIMHANS Convention center\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2015/schedule/graph-algorithms-and-c omputer-vision-9odPE4r4tFV5wENa3iER7d BEGIN:VALARM ACTION:display DESCRIPTION:Graph Algorithms and Computer Vision in Audi 2 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Critical pipe fittings: What every data pipeline requires DTSTART:20150716T041500Z DTEND:20150716T050000Z DTSTAMP:20260423T114800Z UID:session/KrMNxQn2PcpR73AGuSgxXb@hasgeek.com SEQUENCE:2 CATEGORIES:Full Talk,Intermediate CREATED:20150705T131907Z DESCRIPTION:Most organizations leveraging data do so on technologies such as Hadoop\, Spark or Vertica. All these allow organizations to process dat a but nearly always these organizations maintain code base / frameworks et c which the builders use to clean\, process and query this data. While bui lding Starscream (Shopify's dimensional modelling framework on top Spark)\ , we learnt various lessons about numerous building blocks that don't come as part of these technologies yet are critical for smooth functioning and transparency of our data pipeline. The talk aims to provide the audience with these building blocks such as metadata\, incremental builds etc\, the ir use case and how they helped Shopify ship faster.\n[Technology track]\n \n### Speaker bio\n\nYagnik is a software developer at Shopify.\nlinkedin. com/in/yagnik\n GEO:12.943163506799058;77.596340412274 LAST-MODIFIED:20230810T072606Z LOCATION:Audi 1 - NIMHANS Convention center\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2015/schedule/critical-pipe-fittings -what-every-data-pipeline-requires-KrMNxQn2PcpR73AGuSgxXb BEGIN:VALARM ACTION:display DESCRIPTION:Critical pipe fittings: What every data pipeline requires in A udi 1 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Dead Simple Scalability Patterns DTSTART:20150716T050000Z DTEND:20150716T052000Z DTSTAMP:20260423T114800Z UID:session/Gs1LCvcnnonTJBaMZWji9j@hasgeek.com SEQUENCE:2 CATEGORIES:Crisp Talk,Beginner CREATED:20150705T132011Z DESCRIPTION:This talk will highlight development patterns that are easy to catch in code review and go a long way in improving the life of your syst em. For example\,\n\n1. Do not make an **unbounded number** of DB calls in any request\n **Bad Idea**: For each person who retweeted "Ellen's Os car Selfie"\, fetch their avatar from the DB\n2. Do not fetch an **unbound ed amount** of data from the DB\n **Bad Idea**: Fetch all users who re tweeted "Ellen's Oscar Selfie".\n3. Build **timeouts** into every network call made by the system\n **Bad Idea**: Wait forever for this list of RT users to load\, don't render the page until this happens\n\nSlides will list out a large number of "obvious" (and some not-so-obvious) strategies that all distributed systems engineers should follow. For example\,\n\n1. **Data Projections** - Fetch the absolute minimum amount of data required to satisfy a request from the DB\n2. **Simple Profiling** - Count the num ber of DB calls you make to serve a request end-to-end\n3. **Essential Mon itoring** - Measure statistics to determine usefulness. Do you know your c ache hit vs cache miss ratio?\n4. **Awareness of Limits** - What is the v olume throughput limit on Amazon EBS volume?\n\n\nI will also talk about a rchitectural patterns that should be baked in from Day 1. For example\,\n\ n1. Separation of concerns using Message Queues\n2. LRU caching for perman ent\, unchanging data\n3. Version numbers in the schema for feature roll o uts\n\n... and more.\n\n[Technology track]\n\n### Speaker bio\n\nVedang Ma nerikar is a Platform Architect at Helpshift and has helped the Helpshift SDK go from 0 installs to 1 Billion+ installs. Along the way\, he has stay ed up long nights\, refactored multiple systems\, and learned everything i n this talk the hard way. He is also terrible at Markdown.\n GEO:12.943163506799058;77.596340412274 LAST-MODIFIED:20230810T072606Z LOCATION:Audi 1 - NIMHANS Convention center\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2015/schedule/dead-simple-scalabilit y-patterns-Gs1LCvcnnonTJBaMZWji9j BEGIN:VALARM ACTION:display DESCRIPTION:Dead Simple Scalability Patterns in Audi 1 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Running natural language queries against NoSQL schema DTSTART:20150716T050000Z DTEND:20150716T052000Z DTSTAMP:20260423T114800Z UID:session/3PNyT1ySWWBuGurY9FKDmo@hasgeek.com SEQUENCE:2 CATEGORIES:Crisp Talk,Advanced CREATED:20150705T132732Z DESCRIPTION:One of the most common feature of any web application is the " Search". The natural language query module lets users search for what they want with a single text field (rather than a page full of text fields). T he user can type the query that comes to his mind in english and instantly it gets converted into a database query. The solution is designed to iden tify key components and operators within a natural language query and use it against a NoSQL DB to extract the most appropriate records.This session with explain the technical solution in detail.\n\n### Speaker bio\n\n* CT O\, Dexlock Technologies LLP\n* Senior Data Scientist\, Clickagy Inc\n* Sp eaker at MongoDB World Conference on Natural Language Processing\n* Specia lizes in Deeplearning\, NLP and Shallow learning techniques\n GEO:12.943163506799058;77.596340412274 LAST-MODIFIED:20230810T072606Z LOCATION:Audi 2 - NIMHANS Convention center\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2015/schedule/running-natural-langua ge-queries-against-nosql-schema-3PNyT1ySWWBuGurY9FKDmo BEGIN:VALARM ACTION:display DESCRIPTION:Running natural language queries against NoSQL schema in Audi 2 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Morning tea break DTSTART:20150716T052000Z DTEND:20150716T055000Z DTSTAMP:20260423T114800Z UID:session/RVyoGZpuSRfhEGPU1xFaDt@hasgeek.com SEQUENCE:0 CREATED:20150705T132049Z DESCRIPTION:\n LAST-MODIFIED:20150705T132055Z LOCATION:Bangalore ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com BEGIN:VALARM ACTION:display DESCRIPTION:Morning tea break in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Keynote: Data Comes in Shapes DTSTART:20150716T055000Z DTEND:20150716T065000Z DTSTAMP:20260423T114800Z UID:session/NvkoA6ho32nNxYioLasdEP@hasgeek.com SEQUENCE:2 CATEGORIES:Keynote,Beginner CREATED:20150705T132131Z DESCRIPTION:Data comes in shapes. The study of shape is geometry\, in as m any dimensions as you have variables. You can't visualise them all\, but y ou can see in 2D and 3D why the algebraic tools work the way they do: in p articular\, the computer geometry of lines and planes (all the way to big data tools like Support Vector Machines)\, is ‘linear algebra’. I ca n't explain all of linear algebra in 40 minutes\, but I will describe stra tegies for making sense of it without getting confused by all the `a[i][j] ` manipulation.\n\nThink in pictures!\n\n[Keynote - Science talk]\n\n### S peaker bio\n\nProf Tim Poston http://geometeer.com/\n GEO:12.943163506799058;77.596340412274 LAST-MODIFIED:20230810T072606Z LOCATION:Audi 1 - NIMHANS Convention center\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2015/schedule/data-comes-in-shapes-N vkoA6ho32nNxYioLasdEP BEGIN:VALARM ACTION:display DESCRIPTION:Keynote: Data Comes in Shapes in Audi 1 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Keynote: Future patterns in data processing (sponsored) DTSTART:20150716T065000Z DTEND:20150716T073000Z DTSTAMP:20260423T114800Z UID:session/NFvd6NTFobXwXuqm8V7gQC@hasgeek.com SEQUENCE:2 CATEGORIES:Sponsored Keynote,Intermediate CREATED:20150707T073215Z DESCRIPTION:The goodness of an engineering system often is a function of h ow well we understand the pattern of use that the system would see. This t alk focusses on emerging patterns of data consumption/processing\, with th e hope that a better understanding of those will help all of us to create better data systems.\n\n### Speaker bio\n\nI am the CTO of Flipkart.com. M y vision is to build a world-class technology team at Flipkart\, and in tu rn\, do my bit in making India a technology powerhouse. I'm a strong open source proponent\, and a heavy consumer of it at Flipkart. Am excited abou t open sourcing parts of our technology.\n GEO:12.943163506799058;77.596340412274 LAST-MODIFIED:20230810T072606Z LOCATION:Audi 1 - NIMHANS Convention center\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2015/schedule/future-patterns-in-dat a-processing-NFvd6NTFobXwXuqm8V7gQC BEGIN:VALARM ACTION:display DESCRIPTION:Keynote: Future patterns in data processing (sponsored) in Aud i 1 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Lunch Break DTSTART:20150716T073000Z DTEND:20150716T083000Z DTSTAMP:20260423T114800Z UID:session/7SqBYeTsjUsRTkFCtZk2g7@hasgeek.com SEQUENCE:0 CREATED:20150522T142058Z DESCRIPTION:\n LAST-MODIFIED:20150707T073246Z LOCATION:Bangalore ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com BEGIN:VALARM ACTION:display DESCRIPTION:Lunch Break in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Computing languages BOF – Julia and R DTSTART:20150716T083000Z DTEND:20150716T093500Z DTSTAMP:20260423T114800Z UID:session/GpMJ4QjE5sn3WnWo8sguxc@hasgeek.com SEQUENCE:0 CREATED:20150715T073744Z DESCRIPTION:\n GEO:12.943163506799058;77.596340412274 LAST-MODIFIED:20150816T090012Z LOCATION:BoF space - NIMHANS Convention center\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com BEGIN:VALARM ACTION:display DESCRIPTION:Computing languages BOF – Julia and R in BoF space in 5 minu tes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:A review of important results in distributed systems DTSTART:20150716T083000Z DTEND:20150716T091500Z DTSTAMP:20260423T114800Z UID:session/CpeokjyhitSGUHy9u7mdZq@hasgeek.com SEQUENCE:2 CATEGORIES:Full Talk,Intermediate CREATED:20150705T132806Z DESCRIPTION:We will be presenting a summary of some of the key theorms and papers in distributed computing and highlight their impact on system desi gn. We will also go through some of the mitigation methods/workarounds/cle ver hacks on the constraints imposed by these research topics.\n\n### Spea ker bio\n\nI have been a passionate software developer for the last 20+ ye ars. Over that time\, I have worked on a wide variety of platforms and sys tems\, from ZX-Spectrum to large mainframes that run COBOL to embedded sys tems to large distributed systems. OVer the last 10 years\, I have develop ed an interest in machine learning and designing distributed systems. I ha ve presented papers on various events and groups. I think I am more suitab le just because I am passionate about this subject.\n GEO:12.943163506799058;77.596340412274 LAST-MODIFIED:20230810T072606Z LOCATION:Audi 2 - NIMHANS Convention center\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2015/schedule/a-review-of-important- results-in-distributed-systems-CpeokjyhitSGUHy9u7mdZq BEGIN:VALARM ACTION:display DESCRIPTION:A review of important results in distributed systems in Audi 2 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Are these the same pair of shoes? - Matching retail products at sc ale DTSTART:20150716T083000Z DTEND:20150716T091500Z DTSTAMP:20260423T114800Z UID:session/BpRUxz2z1XNCMg56yxEBrm@hasgeek.com SEQUENCE:2 CATEGORIES:Full Talk,Intermediate CREATED:20150705T132947Z DESCRIPTION:Product matching is the problem of resolving product entities across e-commerce sites. This involves a complex sequence of tasks which i nclude - \n1) automatic extraction of key information regions from raw HTM L (for example\, product titles\, UPCs etc.)\n2) categorising products int o a unified taxonomy\n3) semantic parsing of product titles and specificat ions\n4) standardization of attributes such as brands\, colours etc.\, \n5 ) grouping products into clusters of matched products based on a similarit y function or inferencing model. This is a challenging problem because uni que and universally agreed upon identifiers are not always available and p roduct details are noisy and often sparse. So we have to develop contextua l understanding of product specifications\, which are often expressed diff erently by retailers\, merchants\, aggregators etc. \n To scale the matc hing problem to half a billion products\, we also need to prune and bucket effectively while achieving good recall. Matches need to be highly precis e since customers may use them for sensitive tasks such as price compariso n\, competitive analysis and catalog enrichment. We employ an ensemble of online and offline algorithms and models to perform matching at scale for a large number of stores\, categories and brands.\n\n### Speaker bio\n\nNi khil Ketkar leads the data science team at Indix (www.indix.com) which doe s the R&D around product categorization\, standardization\, matching\, sea rch relevance and ranking. He brings along a decade of experience in makin g data-driven decisions and building machine learning models. \nhttps://ww w.linkedin.com/pub/nikhil-ketkar/34/496/a6a\n GEO:12.943163506799058;77.596340412274 LAST-MODIFIED:20230810T072606Z LOCATION:Audi 1 - NIMHANS Convention center\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2015/schedule/are-these-the-same-pai r-of-shoes-matching-retail-products-at-scale-BpRUxz2z1XNCMg56yxEBrm BEGIN:VALARM ACTION:display DESCRIPTION:Are these the same pair of shoes? - Matching retail products a t scale in Audi 1 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:When Apache ZooKeeper is good fit DTSTART:20150716T091500Z DTEND:20150716T093500Z DTSTAMP:20260423T114800Z UID:session/Xsk5V2siMC2BrumrCBSsJA@hasgeek.com SEQUENCE:2 CATEGORIES:Crisp Talk,Intermediate CREATED:20150705T132319Z DESCRIPTION:Apache ZooKeeper is a popular system for coordinating clusters of servers. It is a component that enables the implementation of a number of important tasks for coordination\, such as master election\, group mem bership\, locks. While it is hard to define precisely what ZooKeeper is go od for\, it is not hard to identify use cases in which it is not a good fi t. For example\, it shouldn't be used for bulk data\, file systems and tra ditional databases are much better in such cases. In this presentation\, I discuss typical use cases of ZooKeeper and some other cases in which it i s not a good option.\n\n[Technology track]\n\n### Speaker bio\n\nhttps://i n.linkedin.com/in/rakeshadr\n GEO:12.943163506799058;77.596340412274 LAST-MODIFIED:20230810T072606Z LOCATION:Audi 2 - NIMHANS Convention center\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2015/schedule/when-apache-zookeeper- is-good-fit-Xsk5V2siMC2BrumrCBSsJA BEGIN:VALARM ACTION:display DESCRIPTION:When Apache ZooKeeper is good fit in Audi 2 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Search at Petabyte scale DTSTART:20150716T091500Z DTEND:20150716T093500Z DTSTAMP:20260423T114800Z UID:session/MwH1ko5bfRocqUjGmLP3qc@hasgeek.com SEQUENCE:2 CATEGORIES:Crisp Talk,Intermediate CREATED:20150705T132844Z DESCRIPTION:He is going to talk about the following\n1. Challenges of a la rge fanout distributed system.\n2. Challenges in scaling a multi-tenant sh ared environment.\n3. How do we run search infrastructure at large scale.\ n\n### Speaker bio\n\nAnup heads the Engineering at Sumo Logic India. His team is responsible for Search Engineering at Sumo Logic\, where they work on lot of interesting problems around search performance\, optimization\, reliability.\n\nhttps://in.linkedin.com/in/anupnair\n GEO:12.943163506799058;77.596340412274 LAST-MODIFIED:20230810T072606Z LOCATION:Audi 1 - NIMHANS Convention center\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2015/schedule/search-at-petabyte-sca le-MwH1ko5bfRocqUjGmLP3qc BEGIN:VALARM ACTION:display DESCRIPTION:Search at Petabyte scale in Audi 1 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Social circus – learn juggling\, hooping\, stretches DTSTART:20150716T093500Z DTEND:20150716T104000Z DTSTAMP:20260423T114800Z UID:session/T8VexF7q7Go1HDKP7hFCTV@hasgeek.com SEQUENCE:0 CREATED:20150715T073842Z DESCRIPTION:\n GEO:12.943163506799058;77.596340412274 LAST-MODIFIED:20150813T080406Z LOCATION:BoF space - NIMHANS Convention center\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com BEGIN:VALARM ACTION:display DESCRIPTION:Social circus – learn juggling\, hooping\, stretches in BoF space in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:HawkEye: A Real-Time Anomaly Detection System DTSTART:20150716T093500Z DTEND:20150716T095500Z DTSTAMP:20260423T114800Z UID:session/ASf5ZEwgpsTLQbbi8DWT3B@hasgeek.com SEQUENCE:2 CATEGORIES:Crisp Talk,Beginner CREATED:20150705T132904Z DESCRIPTION:HawkEye is a real-time anomaly detection framework for detecti ng anomalies in IT infrastructure data e.g. CPU\, memory\, response time o f a data center machine. The framework uses a combination of anomaly detec tion techniques to detect local and global anomalies. The system detects s everal types of local anomalies using different anomaly detection techniqu es over a sliding window. We used rigorous data experiments to perform dat a preprocessing and select hyperparameters. We have employed statistics-ba sed techniques to detect local point anomalies and statistical detection t heory (Page’s Test) to detect local contextual anomalies. Time-series mo dels are employed to detect seasonality in the data and detect global anom alies. \nIn this talk\, I will showcase results of the HawkEye system on r eal-data.\n\n### Speaker bio\n\nhttps://in.linkedin.com/in/satnamsinghdata scientist\n GEO:12.943163506799058;77.596340412274 LAST-MODIFIED:20230810T072606Z LOCATION:Audi 1 - NIMHANS Convention center\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2015/schedule/hawkeye-a-real-time-an omaly-detection-system-ASf5ZEwgpsTLQbbi8DWT3B BEGIN:VALARM ACTION:display DESCRIPTION:HawkEye: A Real-Time Anomaly Detection System in Audi 1 in 5 m inutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Exploratory data analysis using Apache Lens and Apache Zeppelin DTSTART:20150716T093500Z DTEND:20150716T095500Z DTSTAMP:20260423T114800Z UID:session/JhVcsHr6dEKaVfm7LDbPjh@hasgeek.com SEQUENCE:2 CATEGORIES:Crisp Talk,Intermediate CREATED:20150705T132401Z DESCRIPTION:Flipkart has multiple data stores catering to diverse requirem ents around flexible questions and response times. At Flipkart\, we use A pache Lens for multi-dimensional queries in a unified way over datasets. T hese datasets are stored in multiple warehouses. Apache Lens aims to ease analytical querying by providing an unified SQL like interface to query ac ross Realtime and historical datasets which could be spanning in across mu ltiple stores. Conceiving data as a cube with hierarchical dimensions lead s to conceptually straightforward operations to facilitate analysis. Integ rating Apache Hive with traditional warehouses and other realtime stores p rovides the opportunity to optimize on the query execution cost while main taining the latency SLAs. \n\nExploratory data analysis is the key ingredi ent behind deriving insights from data. Usually this process requires that multiple people with complementary expertise work together. While most of the query systems provide an interactive shell for data exploration they do not solve for collaboration between multiple people. Apache Zeppelin fi lls this gap.\n\nIn this talk we showcase how Zeppelin and Lens can be use d for collaborative and exploratory data analysis.\n\n[Technology track]\n \n### Speaker bio\n\nBala Nathan is an Architect at Flipkart's Central Dat a platform. He is responsible for flipkart's next generation query and ana lytics platform that will help Analysis and Systems make decisions in real time. He has more than 15 years of experience in the industry. He spent hi s early days in the industry in a couple of startups before he moved on to Oracle and subsequently to Yahoo! where he worked on large scale structur ed and unstructured data ingestion and enrichment platforms. He also does consulting & tech bootstrapping for startups in his spare-time. \n\nPranav Agarwal is an SDE3 at Flipkart's Data platform. At flipkart\, he works on the next generation machine learning platform that uses Apache Zeppelin a nd Lens for exploratory Data analysis. He has more than 15 years of indust ry experience. Prior to Flipkart\, he spent more than decade at Oracle wor king on Oracle Database Search technology that powers Unstructured Text re trieval on RDBMS\, Semantic Search & Content Similarity. He holds a few pa tents on term selection\, document similarity and score boosting.\n GEO:12.943163506799058;77.596340412274 LAST-MODIFIED:20230810T072606Z LOCATION:Audi 2 - NIMHANS Convention center\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2015/schedule/exploratory-data-analy sis-using-apache-lens-and-apache-zeppelin-JhVcsHr6dEKaVfm7LDbPjh BEGIN:VALARM ACTION:display DESCRIPTION:Exploratory data analysis using Apache Lens and Apache Zeppeli n in Audi 2 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Apache Tez - Present and Future DTSTART:20150716T095500Z DTEND:20150716T104000Z DTSTAMP:20260423T114800Z UID:session/XVz2kafGBAh6vx2imXfs4P@hasgeek.com SEQUENCE:2 CATEGORIES:Full Talk,Intermediate CREATED:20150705T132425Z DESCRIPTION:Apache Tez is a framework designed to build data-flow driven p rocessing runtimes. Tez provides a scaffolding and library components that can be used to quickly build scalable and efficient data-flow centric eng ines. This talk will cover the journey of Tez from being a concept in the Apache Incubator to becoming the cornerstone of well-known projects such a s Apache Hive and Apache Pig of the Hadoop ecosystem. I will then move on to the future of Tez on how it is improving to make it easier for data pro cessing applications to be built to run in single-digit seconds and/or to scale to petabytes of data.\n\n[Technology track]\n\n### Speaker bio\n\nRa jesh Balamohan has been working on Hadoop for last couple of years and rec ently has been concentrating on Tez performance at scale.\n GEO:12.943163506799058;77.596340412274 LAST-MODIFIED:20230810T072606Z LOCATION:Audi 2 - NIMHANS Convention center\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2015/schedule/apache-tez-present-and -future-XVz2kafGBAh6vx2imXfs4P BEGIN:VALARM ACTION:display DESCRIPTION:Apache Tez - Present and Future in Audi 2 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Approximate algorithms for summarizing streaming data DTSTART:20150716T095500Z DTEND:20150716T104000Z DTSTAMP:20260423T114800Z UID:session/4fG8bteUZmpz9VM1qB8Mn7@hasgeek.com SEQUENCE:2 CATEGORIES:Full Talk,Intermediate CREATED:20150705T132920Z DESCRIPTION:Approximate algorithms can be used for processing huge streams of incoming data using only a single pass. These algorithms consume finit e amount of memory and cpu cycles. They enable us to maintain summaries wh ich are sufficient to answer expected queries about the data.\n\nTwo such novel algorithms\, finding lots of applications in the industry today are\ n1) Count min sketch (CMS)\n2) HyperLogLog\n\nThis talk aims to:\n1) Provi de a brief introduction to theoritical aspects behind these algorithms \n2 ) How they can be leveraged to summarize unstructured data for practical p urposes.\n3) How to choose the tuning parameters pertinent to your needs.\ n4) Demonstrate how we have used them in Sumologic service.\n\n### Speaker bio\n\nHimadri Sarkar is a Software Engineer at Sumologic India where he is currently working in the are of search performance. Sumo Logic is a clo ud-based log management and analytics service that leverages machine-gener ated big data to deliver real-time IT insights. Search performance team is responsible for delivering all the search related capabilities of the sys tem.\n GEO:12.943163506799058;77.596340412274 LAST-MODIFIED:20230810T072606Z LOCATION:Audi 1 - NIMHANS Convention center\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2015/schedule/approximate-algorithms -for-summarizing-streaming-data-4fG8bteUZmpz9VM1qB8Mn7 BEGIN:VALARM ACTION:display DESCRIPTION:Approximate algorithms for summarizing streaming data in Audi 1 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Evening tea Break DTSTART:20150716T104000Z DTEND:20150716T111500Z DTSTAMP:20260423T114800Z UID:session/De1EtCniqxFug7Hnki6Qoe@hasgeek.com SEQUENCE:0 CREATED:20150522T142648Z DESCRIPTION:\n LAST-MODIFIED:20150707T073450Z LOCATION:Bangalore ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com BEGIN:VALARM ACTION:display DESCRIPTION:Evening tea Break in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Instrumenting your kafka & storm pipeline DTSTART:20150716T111500Z DTEND:20150716T120000Z DTSTAMP:20260423T114800Z UID:session/CoDUZ13EizTWHq2QGmj5XA@hasgeek.com SEQUENCE:2 CATEGORIES:Full Talk,Intermediate CREATED:20150705T132508Z DESCRIPTION:introduction to a production setup that handles billions of ev ents per week through our home-grown apache kafka client\, process the str eam with storm - and then aggregate into postgres. will also share pro's a nd con's of building our own kafka client as opposed to re-using one.\n\n# ## Speaker bio\n\nBosky (@bhaskerkode) leads a product engg team at Helpsh ift & works on erlang\, clojure and golang.\n\nbuilding distributed system s since '06 across edtech\, adtech & mobile in erlang\, clojure & go.\n=> http://in.linkedin.com/in/bhaskerkode & http://slideshare.net/bosky101\n\n Built a kafka producer/micro-service used in production at helpshift\, lay er\, and several other companies.\n=> http://github.com/helpshift/ekaf \n( Recommended by Apache kafka https://cwiki.apache.org/confluence/display/K AFKA/Clients )\n\nUsing storm in production for sentiment analysis\, topic extraction\, naive bayes classification\, etc.\nEager to learn more about best practices in storm deployment/management. Incidentally another part that feeds into this system is in golang\, uses the Shopify kafka producer .\n GEO:12.943163506799058;77.596340412274 LAST-MODIFIED:20230810T072606Z LOCATION:Audi 2 - NIMHANS Convention center\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2015/schedule/instrumenting-your-kaf ka-storm-pipeline-CoDUZ13EizTWHq2QGmj5XA BEGIN:VALARM ACTION:display DESCRIPTION:Instrumenting your kafka & storm pipeline in Audi 2 in 5 minut es TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:CAP Theorem: You don’t need CP\, you don’t want AP\, and you c an’t have CA DTSTART:20150716T111500Z DTEND:20150716T120000Z DTSTAMP:20260423T114800Z UID:session/HYkVbYsEXSMnhXikJy1Gzu@hasgeek.com SEQUENCE:2 CATEGORIES:Full Talk,Intermediate CREATED:20150707T165526Z DESCRIPTION:If you use any distributed database\, you would have surely he ard of the CAP Theorem: "Consistency\, Availability\, Partition tolerance — choose any two!" But CAP theorem is a little more nuanced than that\, and it's impact on how you should choose and operate distributed databases even more so.\n\nConsider these two examples:\n\n1. A distributed key-val ue store where nodes accept all writes but then respond to all read reques ts with a null.\n2. A distributed key-value store where every node rejects all write as well as read requests.\n\nWe can build the above systems usi ng just some bash utilities. The first is an example of an AP system and t he second is an example of a CP system\; in fact\, these examples are more -or-less straight out of the paper that proved the CAP Theorem. Given that such trivial systems can be classified as CP/AP\, is CAP-Theorem even a v iable framework for analysing distributed databases?\n\nTentatively\, the talk will cover:\n\n* intro to CAP theorem (the assertion\; meaning of C\, A\, & P\; proof)\n* examples of trivial CP & AP systems\n* why CA is not meaningful\, CP is not viable\, AP is not useful\n* alternatives to CAP th eorem\n* analysis of a few real-world databases\n\n### Speaker bio\n\nSidd hartha is an Architect at Flipkart\, presently working on the company's ce ntral Data Platform. He had earlier built and operated Flipkart.com's sear ch system and offers platform for a couple of years each. Through these an d other systems he has been exposed to several distributed data systems wi th varying needs of scale\, consistency\, & availability. Siddhartha's are as of interest include distributed systems\, databases\, and information r etrieval.\n GEO:12.943163506799058;77.596340412274 LAST-MODIFIED:20230810T072606Z LOCATION:Audi 1 - NIMHANS Convention center\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2015/schedule/cap-theorem-you-dont-n eed-cp-you-dont-want-ap-and-you-cant-have-ca-HYkVbYsEXSMnhXikJy1Gzu BEGIN:VALARM ACTION:display DESCRIPTION:CAP Theorem: You don’t need CP\, you don’t want AP\, and y ou can’t have CA in Audi 1 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Making a contextual recommendation engine using Python and Deep Le arning at ParallelDots DTSTART:20150716T120000Z DTEND:20150716T122000Z DTSTAMP:20260423T114800Z UID:session/UuybGZMaSpb5MWe13f8VUi@hasgeek.com SEQUENCE:2 CATEGORIES:Crisp Talk,Beginner CREATED:20150705T133926Z DESCRIPTION:Making a Recommendation Engine at ParallelDots (4 minutes)\n\n a. Why normal full-text search will not work: The problem of incorrect tag ging and slow search queries.\n\nb. ParallelDots' MVP with Topic Models: I ssues with accuracy and scaling.\n\nc. Decision to use Deep Learning and a ims of the new architecture (Not enough funds for distributed system\, sea rch related posts from millions of documents in reasonable time)\n\n Ba sics of Deep Learning ( 12 min )\n\na. Deep Neural Networks\n\nb. Types of Deep Neural Networks. Convolutional\, DBNs\, Recurrent and Recursive. How do they differ in structure\, types of neurons and training.\n\nc. Backpr opogation and its variants\n\nd. Features of various Deep Learning librari es in Python. Theano\, PyLearn2\, theanets\, lasagne\, kayak\, pybrain.\n\ n Deep Learning in NLP ( 10 min )\n\na. Solving problem of high-dimensi onality using word embeddings.\n\nb. Common approaches to word embedding.\ n\nc. Modelling language as a series of characters using Recurrent Neural Networks .\n\nc. Models we use : Named Entity Recognition with Neural Nets \n\nd. Models we use: Combining word embeddings using heuristics and recur sive neural networks.\n\n Search Engine ( 10 min )\n\na. Using Search D ata Structures to convert search related posts operations from O(n) to O(l og(n))\n\nb. Space Partitioning Trees : Search for nearest Neighbours. Exa mples of such trees: KD-Tree / Ball Tree / VP Tree\n\nc. Why we chose VP T ree ? What libraries to use to code up in Python ?\n\nd. Parallelization. Data Parallel Python's multiprocessing parallelization not the best\, work ing towards a shared memory parallel version.\n\n Scaling up system ( 2 min )\n\na. Hacks to scale up recommendations.\n\nb. Using golang's chann els to unique requests.\n\n Q&A ( 2 min )\n\n### Speaker bio\n\nMuktabh is one of co-founders of ParallelDots. He handles the Data Science and So ftware Architecture at the startup. Previously\, He has worked at Opera So lutions and as consultant Data Scientist\, wherein he helped solving many data based problems in healthcare\, internet\, procurement\,retail and per sonal finance . He has a degree in Information Systems from BITS Pilani\, Pilani. Social handles: in.linkedin.com/in/muktabh / @muktabh\n GEO:12.943163506799058;77.596340412274 LAST-MODIFIED:20230810T072606Z LOCATION:Audi 2 - NIMHANS Convention center\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2015/schedule/making-a-contextual-re commendation-engine-using-python-and-deep-learning-at-paralleldots-UuybGZM aSpb5MWe13f8VUi BEGIN:VALARM ACTION:display DESCRIPTION:Making a contextual recommendation engine using Python and Dee p Learning at ParallelDots in Audi 2 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Joining data streams at scale for fun and profit DTSTART:20150716T120000Z DTEND:20150716T122000Z DTSTAMP:20260423T114800Z UID:session/WRR7SJzyUnJLsb8U3wXMEx@hasgeek.com SEQUENCE:2 CATEGORIES:Crisp Talk,Beginner CREATED:20150705T133739Z DESCRIPTION:Real-time data streams are everywhere\, which is not really su rprising considering how easy it has gotten to generate them. For example\ , applications can write their hearts out to a Kafka cluster\, logs can be streamed out via Logstash\, a Change Data Capture (CDC) system can be dep loyed to to turn a database’s write-ahead log into a stream\, etc.\n\nSt ream processing systems such as Apache Storm have become quite popular for analysing such data streams\; they are used to power real-time analytical dashboards as well as other data-driven products such as recommendations\ , trending topics etc.\n\nBut a lot more value could be derived by joining multiple data streams in real-time — when it comes to data\, the value of the whole is much greater than the parts. For example\, while data stre ams of search queries and product result clicks are useful by themselves\, joining them allows us to derive metrics such as clickthrough-rate. But d oing this is wrought with several challenges:\n\n* Data streams could be c oming in — or be processed with — different lags i.e. one stream could have data as recent as a few seconds ago while the latest data in another stream is from several minutes ago. This can occur due to buffering\, cro ss datacenter replication etc.\n* Data streams could have data that is com ing out-of-order i.e. while most of the latest few events in a data stream are from a few seconds ago\, a small percentage could be from several min utes ago. This can also occur due to buffering\, cross datacenter replicat ion etc.\n* Data streams could be very large\, to the tune of tens to hund reds of thousands of events per second.\n* Data streams could represent up dates to some mutable state such as updates to a User Profile\, an Order\, Product Information etc. While this is not a big challenge by itself\, it becomes much more complex to deal with in the context of the above challe nges.\n\nStream processing systems such as Apache Storm do not provide any out of the box ways of expressing such complex joins\, let alone taking c are of the complications listed above. At Flipkart\, we built a framework on top of Storm to do the same\; this framework has been used for building hundreds of stream processing pipelines joining Kafka data streams.\n\nIn this talk\, we’ll describe the design patterns that the framework imple ments\, how we arrived at them and the lessons we learnt along the way.\n\ n### Speaker bio\n\nAniruddha works as a Software Engineer at Flipkart. He is the lead developer for the stream processing framework of the company' s central Data Platform. In this role Aniruddha was responsible for buildi ng and operationalising the design patterns being described in this talk.\ n\nSiddhartha Reddy is an Architect at Flipkart. He works with Aniruddha o n the stream processing framework described here.\n GEO:12.943163506799058;77.596340412274 LAST-MODIFIED:20230810T072606Z LOCATION:Audi 1 - NIMHANS Convention center\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2015/schedule/joining-data-streams-a t-scale-for-fun-and-profit-WRR7SJzyUnJLsb8U3wXMEx BEGIN:VALARM ACTION:display DESCRIPTION:Joining data streams at scale for fun and profit in Audi 1 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Revolutionizing travel with ML & Analytics – An insight into bus iness optimization using Machine Learning and Advanced Analytics DTSTART:20150716T122000Z DTEND:20150716T130500Z DTSTAMP:20260423T114800Z UID:session/WG3da4w33LaSe9kCAgqtN1@hasgeek.com SEQUENCE:2 CATEGORIES:Full Talk,Intermediate CREATED:20150708T023448Z DESCRIPTION:Big data technology is at the heart of our business. Our Trave l sites generate about 1 TB of data per day which is stored in HDFS. This data is at the heart of machine learning\, advanced analytics and visual analytics. We go over a few case studies which have had a great impact on travel experience for our users and the technology behind them.\n\n1. Usag e of machine learning to provide personalized sorting of search results wh ich in turn increased the propensity to buy.\n2. Collecting and analyzing Site experimentation data.\n3. Marketing Channel Optimization and Campaign effectiveness\n4. Data platform that enables all of our use cases\n\n### Speaker bio\n\nRaghu Kashyap serves as the Sr Director of Technology at Or bitz Worldwide. Raghu possesses an extensive background in technology and travel and leads the Data Infrastructure functions at Orbitz Worldwide. Ra ghu’s team leverages Hadoop extensively to augment traditional BI/DW and unlock the joy of travel for its customers and drive business performance . \n\nRaghu also heads the Bangalore development center of Orbitz which fo cuses on application development and data science and EDW platform enablem ent\n GEO:12.943163506799058;77.596340412274 LAST-MODIFIED:20230810T072606Z LOCATION:Audi 1 - NIMHANS Convention center\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2015/schedule/revolutionizing-travel -with-ml-analytics-an-insight-into-business-optimization-using-machine-lea rning-and-advanced-analytics-WG3da4w33LaSe9kCAgqtN1 BEGIN:VALARM ACTION:display DESCRIPTION:Revolutionizing travel with ML & Analytics – An insight into business optimization using Machine Learning and Advanced Analytics in Au di 1 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Check-in DTSTART:20150717T030000Z DTEND:20150717T041500Z DTSTAMP:20260423T114800Z UID:session/M7F8rFfVNz6UUxpXQ4gbpf@hasgeek.com SEQUENCE:0 CREATED:20150616T082720Z DESCRIPTION:\n LAST-MODIFIED:20150705T133253Z LOCATION:Bangalore ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com BEGIN:VALARM ACTION:display DESCRIPTION:Check-in in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Deep Learning for Natural Language Processing DTSTART:20150717T041500Z DTEND:20150717T050000Z DTSTAMP:20260423T114800Z UID:session/8yxhdTMeHz62Kad51oEAeA@hasgeek.com SEQUENCE:2 CATEGORIES:Full Talk,Intermediate CREATED:20150706T050308Z DESCRIPTION:Much of the Text Mining needed in real-life boils down to Text Classification: be it prioritising e-mails received by Customer Care\, ca tegorising Tweets aired towards an Organisation\, measuring impact of Prom otions in Social Media\, and (Aspect based) Sentiment Analysis of Reviews. These techniques can not only help gauge the customer's feedback\, but al so can help in providing users a better experience. \n \nTraditional solut ions focused on heavy domain-specific Feature Engineering\, and thats exac tly where Deep Learning sounds promising!\n \nWe will depict our foray int o Deep Learning with these classes of Applications in mind. Specifically\, we will describe how we tamed Deep Convolutional Neural Network\, most co mmonly applied to Computer Vision\, to help classify (short) texts\, attai ning near-state-of-the-art results on several SemEval tasks consistently\, and a few tasks of importance to Flipkart.\n\nIn this talk\, we plan to c over the following:\n\n- Basics of Deep Learning as applied to NLP: Word E mbeddings and its compositions a la Recursive Neural Networks\, Convolutio nal Neural Networks\, and Recurrent Neural Networks.\n\n- New Experimental results on an array of SemEval / Flipkart's internal tasks: e.g. Tweet Cl assification and Sentiment Analysis. (As an example we achieved 95% accura cy in binary sentiment classification task on our datasets - up from 85% b y statistical models)\n\n- Share some of the learnings we have had while d eploying these in Flipkart!\n\nHere is a mindmap explaining the flow of co ntent and key takeawys for the audience: https://atlas.mindmup.com/2015/06 /4cbcef50fa6901327cdf06dfaff79cf0/deep_learning_for_natural_language_proce /index.html\n\n### Speaker bio\n\nDevashish is a software developer at Fli pkart. His primary work is in the area of deriving insights from unstructu red data. His previous works include Review Summarization at Flipkart\, Us er Insights Platform - a system which enables us to dig into 100s of TB of data to find user insights. He is currently working on sentiment analysis and aspect extraction from Social Media.\n\nPrerana Singhal is currently pursuing Masters of Technology in the department of Computer Science from IIT Bombay. She is doing her research work in the field Natural Language Processing under Professor Pushpak Bhattacharya\, the principle area of in terest being Sentiment Analysis and Deep Learning.\n GEO:12.943163506799058;77.596340412274 LAST-MODIFIED:20230810T072606Z LOCATION:Audi 1 - NIMHANS Convention center\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2015/schedule/deep-learning-for-natu ral-language-processing-8yxhdTMeHz62Kad51oEAeA BEGIN:VALARM ACTION:display DESCRIPTION:Deep Learning for Natural Language Processing in Audi 1 in 5 m inutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:The many ways of parallel computing with Julia DTSTART:20150717T041500Z DTEND:20150717T050000Z DTSTAMP:20260423T114800Z UID:session/6daBRGgHkt1WayT6fWV2jN@hasgeek.com SEQUENCE:2 CATEGORIES:Full Talk,Beginner CREATED:20150705T133330Z DESCRIPTION:Julia is a high performance dynamic language\, primarily desig ned for technical computing\, but increasingly seeing applications in a wi de variety of domains.\n\nThis talk will provide an overview of parallel c omputing in Julia. It will start with an introduction to using built-in Ju lia primitives for parallel processing\, such as pmap\, @parallel\, remote call\, spawn\, fetch\, etc. Based on this low-level primitives\, shared ar rays and distributed arrays have been built. We will try some Parallel Lin ear Algebra using packages such as ScaLapack along with some MPI programmi ng. We will also look at the possibilities of data processing with data lo aded from the Hadoop file system (HDFS) and/or S3. We will also preview th e upcoming multi-threading capabilities in Julia.\n\nNot only will we show how the compute can be efficiently carried out in Julia\, but using tools such as Escher\, we will also show how the findings can be beautifully pa ckaged for presentation.\n\n[Technology track]\n\n### Speaker bio\n\nhttps ://in.linkedin.com/in/viralbshah\n GEO:12.943163506799058;77.596340412274 LAST-MODIFIED:20230810T072606Z LOCATION:Audi 2 - NIMHANS Convention center\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2015/schedule/the-many-ways-of-paral lel-computing-with-julia-6daBRGgHkt1WayT6fWV2jN BEGIN:VALARM ACTION:display DESCRIPTION:The many ways of parallel computing with Julia in Audi 2 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Using Modes for Time Series Classification DTSTART:20150717T050000Z DTEND:20150717T052000Z DTSTAMP:20260423T114800Z UID:session/4oiCe7GgKrBx9qPdGqRG5Z@hasgeek.com SEQUENCE:2 CATEGORIES:Crisp Talk,Beginner CREATED:20150706T050314Z DESCRIPTION:The time series that appear in IT monitoring environments are not always amenable to ARIMA or related methods. I have been exploring the mode as a statistic from which to build a signature for a time series. If time permits I will show how standard ML methods can be applied to these signatures for classification.\n\n### Speaker bio\n\nWas almost a mathemat ician\, then moved to HFT as a C++ / Perl developer\, nowadays C#\, Python and JavaScript. Just enough to do analytics on IT monitoring data using p andas numpy etc.\n GEO:12.943163506799058;77.596340412274 LAST-MODIFIED:20230810T072606Z LOCATION:Audi 1 - NIMHANS Convention center\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2015/schedule/using-modes-for-time-s eries-classification-4oiCe7GgKrBx9qPdGqRG5Z BEGIN:VALARM ACTION:display DESCRIPTION:Using Modes for Time Series Classification in Audi 1 in 5 minu tes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Harnessing the power of the Erlang VM at Housing DTSTART:20150717T050000Z DTEND:20150717T052000Z DTSTAMP:20260423T114800Z UID:session/B2wbk51KwMF3wZYqBdaefy@hasgeek.com SEQUENCE:2 CATEGORIES:Crisp Talk,Intermediate CREATED:20150705T133406Z DESCRIPTION:We will discuss the following three use cases as examples of w here we used the Erlang VM to overcome limitations of speed and memory in our core infrastructure\n\n1. Ruby APIs\, implementing search with Elastic search\, have been the backbone of our realtime search infrastructure. Aft er having experienced a few server side scalability bottlenecks due to ine fficient Unicorn workers\, we decided to port some of the feasible and cri tical code to Elixir which exploits the soft real-time capabilities of the Erlang VM. In a Service Oriented Architecture centric environment\, vario us Ruby APIs doing IO on network fitted a great use case for the Erlang VM . \n\n2. CouchDB is another software written in Erlang. It plays a signifi cant role in storage of custom JSON data and enable real-time streaming se arch through CouchDB Views. We injected a Elixir evaluation runtime in Co uchDB to allow us to write views in Elixir\, which results in performant V iew generation times while also assuring developer productivity. \n\n3. R abbitMQ has been a critical component for our backend infrastructure . The use of topic exchanges can enable the source logic and the consumer logic to be decoupled\, but in the quest to provide generic flexibility in rout ing\, inefficiency propagates to the consumers written in different langua ges which filter out the messages based on a custom logic.\n\n We propo se to insert an evaluation engine into RabbitMQ wherein consumers define c ustom logic based code in Erlang/Elixir and Javascript in order to filter messages and deliver only relevant messages. Moreover\, to combat abnormal memory usage issues in RabbitMQ\, we experimented with a few VM strategie s and discovered some novel ways of keeping RabbitMQ memory usage consiste nt without sacrificing significant performance.\n \n [Technology track ]\n\n### Speaker bio\n\nAbhijit Pratap Singh is a Software Developer at Ho using.com. He has been working at Housing.com for the past two years. He h as played an active part in developing the product backend for Housing.com .\n\nPranav Rao is a Software Developer at Housing.com. He was the primary developer behind the geo-services backend at Housing.com working on searc h and storage of other geo-entities.\n GEO:12.943163506799058;77.596340412274 LAST-MODIFIED:20230810T072606Z LOCATION:Audi 2 - NIMHANS Convention center\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2015/schedule/harnessing-the-power-o f-the-erlang-vm-at-housing-B2wbk51KwMF3wZYqBdaefy BEGIN:VALARM ACTION:display DESCRIPTION:Harnessing the power of the Erlang VM at Housing in Audi 2 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Morning tea break DTSTART:20150717T052000Z DTEND:20150717T055000Z DTSTAMP:20260423T114800Z UID:session/5MATfvCg623nv9y1L886RV@hasgeek.com SEQUENCE:0 CREATED:20150616T082919Z DESCRIPTION:\n LAST-MODIFIED:20150705T133432Z LOCATION:Bangalore ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com BEGIN:VALARM ACTION:display DESCRIPTION:Morning tea break in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Keynote: "Thinking Machines" DTSTART:20150717T055000Z DTEND:20150717T065000Z DTSTAMP:20260423T114800Z UID:session/Da2etG9DiyBqh6Ky2tTqUh@hasgeek.com SEQUENCE:2 CATEGORIES:Keynote,Advanced CREATED:20150708T022047Z DESCRIPTION:Ever since the dawn of civilization humans have been trying to "extend" themselves through "machines" - tools to hunt animals stronger t han us\, telescopes to see farther than our eyes\, cranes to lift more tha n our hands\, cars to move faster than our legs\, and computers to store a nd process more information than ever before. The epitome of this "extensi on"\, the dream (or nightmare) of Artificial Intelligence\, is to create a "Thinking Machine" that replicates or even surpasses our brain's ability to:\n\n(a) "Understand" the data it receives\, \n\n(b) "Learn" composition al and causal structures in it\, \n\n(c) "Reason" over these structures\, and even \n\n(d) "Create" new data using the "model" of the world it has l earnt. \n\nSuch a machine could\, for example\, have a human like conversa tion\, solve complex problems\, predict the next scene in a video\, find a cure for cancer\, prove Fermat's last theorem\, or even create art or poe try some day. Advances made in machine learning\, natural language underst anding\, speech recognition and synthesis\, computer vision\, and deep lea rning have brought us to the brink of such a possibility! We can already s ee glimpses of such "thinking machines" - the self driving cars\, the jeop ardy winning Watson\, and the recent "paintings" created from deep vision networks (http://googleresearch.blogspot.in/2015/06/inceptionism-going-dee per-into-neural.html).\n\nIn this talk\, we will explore both the philosop hical and algorithmic aspects of building such a "Thinking Machine". More specifically\, we will explore some hypothese on how we:\n\n(a) learn\, re present\, and disambiguate senses of words without knowing the grammar of a language?\n\n(b) analyze knowledge from langauge and synthesize language from knowledge?\n\n(c) reason over this knowledge to solve complex proble ms and innovate new ideas?\n\n(d) can build better machine translation and conversation systems using some of these building blocks?\n\n### Speaker bio\n\nhttps://www.linkedin.com/in/shaileshk\n GEO:12.943163506799058;77.596340412274 LAST-MODIFIED:20230810T072606Z LOCATION:Audi 1 - NIMHANS Convention center\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2015/schedule/thinking-machines-Da2e tG9DiyBqh6Ky2tTqUh BEGIN:VALARM ACTION:display DESCRIPTION:Keynote: "Thinking Machines" in Audi 1 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Deploying Batch and Streaming Architectures on AWS (sponsored) DTSTART:20150717T065000Z DTEND:20150717T073000Z DTSTAMP:20260423T114800Z UID:session/S35bd11HyNY1xFboQaZebj@hasgeek.com SEQUENCE:2 CATEGORIES:Sponsored,Intermediate CREATED:20150705T133603Z DESCRIPTION:One of the biggest challenges organizations face when designin g Big Data platforms is analyzing historical\, batch and streaming data us ing the same architecture. \n\nThis session will illustrate how to use AWS Big Data services such as Amazon Elastic MapReduce\, Amazon Kinesis\, Ama zon Redshift and others to build a scalable\, fault-tolerant and multi-lay ered processing system which includes the ability to analyze streaming dat a by comparing it against historical data in near real-time.\n\n### Speake r bio\n\nRussell Nash is a Solutions Architect with Amazon focused on data analytics.\n\nHe works with customers to derive maximum value and perform ance from using the Amazon data analytics services.\n\nRussell has over 20 years experience in the IT industry and the majority of that time was spe nt working with database and parallel technologies designed for large scal e analysis.\n\nHe is passionate about applying these technologies to busin ess problems in order to return value and insight to organizations.\n GEO:12.943163506799058;77.596340412274 LAST-MODIFIED:20230810T072606Z LOCATION:Audi 2 - NIMHANS Convention center\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2015/schedule/deploying-batch-and-st reaming-architectures-on-aws-S35bd11HyNY1xFboQaZebj BEGIN:VALARM ACTION:display DESCRIPTION:Deploying Batch and Streaming Architectures on AWS (sponsored) in Audi 2 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Building a E-commerce search engine: Challenges\, insights and app roaches (sponsored) DTSTART:20150717T065000Z DTEND:20150717T073000Z DTSTAMP:20260423T114800Z UID:session/Wok3p2eARBiAC4y7BNhbFx@hasgeek.com SEQUENCE:2 CATEGORIES:Sponsored,Beginner CREATED:20150705T134137Z DESCRIPTION:In this talk\, we go into the details of what goes behind the scenes in building a state-of-the-art E-Commerce search engine. Over the l ast decade\, there is a huge surge in the amounts of (text and even non-te xt) data that is not only being collected and stored but also being access ed - e.g. web documents (e.g Google)\, user activities (e.g. facebook\, pi nterest\, ads) and search engine applications have become one critical way to access this data on demand. While scale of the data is a inherent prob lem to be addressed while building search engines\, e-commerce application s present its own set of challenges that make building an e-commerce searc h engine unique and markedly different from traditional search application s. In this talk\, we will analyze the challenges\, gather insights from th is problem space and go over a few approaches on how to go about building a world-class product search engine\n\n### Speaker bio\n\nVinodh Kumar is the CTO and M.D of BloomReach India. At BloomReach\, Vinodh Kumar has been leading the engineering and development of BloomReach's e-commerce search engine product called SNAP (Search\, Navigation and Personalization). Pri or to BloomReach\, Vinodh Kumar worked at Google\, where he was the man be hind the Google News ranking algorithms\, where he served as the Tech Lead of the Google News Quality team. Vinodh Kumar has a bachelors from Anna U niversity and Masters in Computer Science from IISc. He was also the All I ndia topper in GATE Computer science in 1999.\n GEO:12.943163506799058;77.596340412274 LAST-MODIFIED:20230810T072606Z LOCATION:Audi 1 - NIMHANS Convention center\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2015/schedule/building-a-e-commerce- search-engine-challenges-insights-and-approaches-Wok3p2eARBiAC4y7BNhbFx BEGIN:VALARM ACTION:display DESCRIPTION:Building a E-commerce search engine: Challenges\, insights and approaches (sponsored) in Audi 1 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Lunch DTSTART:20150717T073000Z DTEND:20150717T083000Z DTSTAMP:20260423T114800Z UID:session/EhrVEzCnv47p8jvqab8dKV@hasgeek.com SEQUENCE:0 CREATED:20150616T083036Z DESCRIPTION:\n LAST-MODIFIED:20150707T074125Z LOCATION:Bangalore ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com BEGIN:VALARM ACTION:display DESCRIPTION:Lunch in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Apache Spark BOF DTSTART:20150717T083000Z DTEND:20150717T093500Z DTSTAMP:20260423T114800Z UID:session/2S41XsQ4Y9yBTuu49CHius@hasgeek.com SEQUENCE:0 CREATED:20150715T074035Z DESCRIPTION:\n GEO:12.943163506799058;77.596340412274 LAST-MODIFIED:20150715T083556Z LOCATION:BoF space - NIMHANS Convention center\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com BEGIN:VALARM ACTION:display DESCRIPTION:Apache Spark BOF in BoF space in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Call me maybe: Jepsen and flaky networks DTSTART:20150717T083000Z DTEND:20150717T091500Z DTSTAMP:20260423T114800Z UID:session/spKJQfqccuNMbuYbC2kqC@hasgeek.com SEQUENCE:2 CATEGORIES:Full Talk,Advanced CREATED:20150705T133515Z DESCRIPTION:In the big data world\, our data stores communicate over an as ynchronous\, unreliable network to provide a facade of consistency. Howeve r\, to really understand the guarantees of these systems\, we must underst and the realities of networks and test our data stores against them.\n\nJe psen is a tool which simulates network partitions in data stores and helps us understand the guarantees of our systems and its failure modes. In thi s talk\, I will help you understand why you should care about network part itions and how can we test datastores against partitions using Jepsen. I w ill explain what Jepsen is and how it works and the kind of tests it lets you create. We will try to understand the subtleties of distributed consen sus\, the CAP theorem and demonstrate how different data stores such as Mo ngoDB\, Cassandra\, Elastic and Solr behave under network partitions. Fina lly\, I will describe the results of the tests I wrote using Jepsen for Ap ache Solr and discuss the kinds of rare failures which were found by this excellent tool.\n\n### Speaker bio\n\nI am a committer on Apache Lucene/So lr since 2008 as well as a member of the Lucene/Solr project management co mmittee. I’ve worked at AOL for five years on vertical search\, content mangement systems\, social/community platforms and anti-spam systems as we ll as AOL WebMail’s Inbox Search system which uses a highly customized v ersion of Apache Solr to service tens of millions of users and more than a billion index/search operations a day. I currently work at Lucidworks Inc . on Apache Solr and Lucidworks Search mostly on the SolrCloud side of thi ngs. I also help organize the Bangalore Apache Solr/Lucene Meetup Group wh ich has 450+ members and holds regular meetings of people interested in Lu cene\, Solr and search in general.\n\nhttps://twitter.com/shalinmangar\n\n https://in.linkedin.com/in/shalinsmangar\n\nhttp://shal.in/\n GEO:12.943163506799058;77.596340412274 LAST-MODIFIED:20230810T072606Z LOCATION:Audi 2 - NIMHANS Convention center\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2015/schedule/call-me-maybe-jepsen-a nd-flaky-networks-spKJQfqccuNMbuYbC2kqC BEGIN:VALARM ACTION:display DESCRIPTION:Call me maybe: Jepsen and flaky networks in Audi 2 in 5 minute s TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Keeping Moore's law alive: Neuromorphic computing DTSTART:20150717T083000Z DTEND:20150717T091500Z DTSTAMP:20260423T114800Z UID:session/Xvu2sRfvQYZtjkgd7wQfk6@hasgeek.com SEQUENCE:2 CATEGORIES:Full Talk,Beginner CREATED:20150705T134040Z DESCRIPTION:Moore's law is a term coined by Carver Mead\, a Caltech profes sor who is also the father of Neuromorphic Engineering. It refers to the o bservation\, now more hope than reality\, that advances in technology will allow a doubling of compute capability in silicon every 18 months. Recent advances in the use of highly parallel compute methods\, that are loosely based on neural systems in our brain\, are changing how compute is accomp lished. These techniques\, collectively termed deep learning networks\, bu rst onto to the world because of one reason: the ability to perform lots o f parallel computations on graphics cards. However\, it is in truly custom hardware\, such as that pioneered by the Neuromorphic community that we w ill find the salvation of Moore's law. When we blend powerful compute tech niques with custom silicon architectures\, we can keep the hope alive of c ontinuing to double the compute capability of the world.\n\nIf you are in the space of deep learning or have heard about how GPUs have revolutionali zed high performance computing\, this talk will take you to the extreme bl eeding edge of that world.\n\n### Speaker bio\n\nThe speaker was one of th e creators of Neurogrid\, a system built in Stanford that until recently w as the largest Neuromorphic system in the world. He is also the CTO and Fo under of Mad Street Den\, a computer vision and AI startup based out of Ch ennai.\n GEO:12.943163506799058;77.596340412274 LAST-MODIFIED:20230810T072606Z LOCATION:Audi 1 - NIMHANS Convention center\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2015/schedule/keeping-moores-law-ali ve-neuromorphic-computing-Xvu2sRfvQYZtjkgd7wQfk6 BEGIN:VALARM ACTION:display DESCRIPTION:Keeping Moore's law alive: Neuromorphic computing in Audi 1 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:POC: How to slice\, dice & search billions of users events in seco nds (from scratch) DTSTART:20150717T091500Z DTEND:20150717T093500Z DTSTAMP:20260423T114800Z UID:session/GZsbp42a8F1PBw21Pu6qht@hasgeek.com SEQUENCE:2 CATEGORIES:Crisp Talk,Beginner CREATED:20150705T133534Z DESCRIPTION:Supported actions:\n\nCouting the number of users performing a n event(cardinality)\nCounting the number of users performing a combinatio n of 1000 events\nHow to support searching events across time ranges (to s uit your case)\n\n1) bits to be compared\n2) bits intersection\n3) time fo r running intersection\n4) time for counting intersection\n5) hacks & tric ks when working with bits\, popcount & bitmaps\n\nFirst version up in two days. Will show you how to get it up and running for your internal team as well.\n\n### Speaker bio\n\nBosky (@bhaskerkode) leads a product engg tea m at Helpshift & works on erlang\, clojure and golang.\n\nbuilding distrib uted systems since ‘06 across edtech\, adtech & mobile in erlang\, cloju re & go.\n\nmore talks at http://slideshare.net/bosky101\n\nmore about bos ky at http://in.linkedin.com/in/bhaskerkode\n GEO:12.943163506799058;77.596340412274 LAST-MODIFIED:20230810T072606Z LOCATION:Audi 2 - NIMHANS Convention center\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2015/schedule/poc-how-to-slice-dice- search-billions-of-users-events-in-seconds-from-scratch-GZsbp42a8F1PBw21Pu 6qht BEGIN:VALARM ACTION:display DESCRIPTION:POC: How to slice\, dice & search billions of users events in seconds (from scratch) in Audi 2 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Hardware Accelerated Big Data Processing DTSTART:20150717T091500Z DTEND:20150717T093500Z DTSTAMP:20260423T114800Z UID:session/6JwbSmDpRtCbhocv1j4fPi@hasgeek.com SEQUENCE:2 CATEGORIES:Crisp Talk,Intermediate CREATED:20150705T134106Z DESCRIPTION:Use of FPGAs (Field Programmable Gate Arrays) for big data pro cessing has recently started rising:\n\n- *Intel Buying [FPGA] Chipmaker A ltera For $16.7 Billion* (Forbes\, 1-6-15)\n- *Microsoft Supercharges Bing Search with [FPGA] Programmable Chips* (Wired\, 6-16-14)\n\nWhats the big deal? Proposed talk aims to elucidate. Outline:\n\n- Quick\, intuitive\, technical introduction to FPGAs and their key benefits over software for b ig data\n- Computation structures that are useful for big data processing and can be efficiently implemented on FPGAs\n- Pros/cons\, challenges/oppo rtunities in FPGA use for big data\n- Concrete use-case of FPGA accelerate d XML data processing using novel tree automata based compact implementati on that is multiple times faster than a Xeon processor\n\n### Speaker bio\ n\nReetinder Sidhu currently leads project LEAP (Logic Engines for Acceler ated Processing) in Tech Mahindra's Intrapreneurship Program. He has a PhD (EE\, FPGA area) from University of Southern California and an MS (comput er engg.) from Indian Institute of Science.\n\nHe is an active FOSS user a nd enthusiast since the days Linux came on floppies and installing gcc inv olved compiling the compiler...\n GEO:12.943163506799058;77.596340412274 LAST-MODIFIED:20230810T072606Z LOCATION:Audi 1 - NIMHANS Convention center\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2015/schedule/hardware-accelerated-b ig-data-processing-6JwbSmDpRtCbhocv1j4fPi BEGIN:VALARM ACTION:display DESCRIPTION:Hardware Accelerated Big Data Processing in Audi 1 in 5 minute s TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Building tiered data stores using Aesop to bridge SQL and NoSQL sy stems DTSTART:20150717T093500Z DTEND:20150717T102000Z DTSTAMP:20260423T114800Z UID:session/LmE2pAiPNSHjY7gTpN7P8R@hasgeek.com SEQUENCE:2 CATEGORIES:Full Talk,Intermediate CREATED:20150705T133719Z DESCRIPTION:Large scale internet systems often use a combination of relati onal (SQL) and non-relational (NoSQL) data stores. Contrary to product cla ims\, it is hard to find a single data store that meets common read-write patterns of on-line applications. Different databases try to optimize for specific workload patterns and data durability\, consistency guarantees - use Memory buffer pools\, Write-ahead logs\, optimize for Flash storage et c. These data stores are not operated in isolation and need to share data and updates on it - for e.g. a high performance memory based KV data cache might need to be updated when data in the source-of-truth RDBMS or Column ar database changes. \n\nThis talk discusses general approaches to Change Data Propagation and specific implementation details of Flipkart's open-so urce project : [Aesop](https://github.com/Flipkart/aesop)\, including some of its live deployments. It covers capabilities suitable for single node deployment and also scale to multi-node partitioned clusters that process data concurrently at high throughput.\n\nAesop scales by partitioning the data stream and coordinates across subscription nodes using Zookeeper. It provides atleast-once delivery guarantees and timeline ordered data update s.\n\nAesop is used at scale in business critical systems - the multi-tier ed payments data store\, the user wishlist system and streaming facts to d ata analysis platform. A number of upcoming adopters include the Promotion s and Warehousing systems backend data stores. Aesop has been used success fully to move millions of data records between MySQL\, HBase\, Redis\, Kaf ka and Elastic Search clusters. \n\nAesop shares common design approach an d technologies with \n[Facebook Wormhole system](https://www.usenix.org/sy stem/files/conference/nsdi15/nsdi15-paper-sharma.pdf)\n\nCome attend this talk if you are evaluating data store(s) for your large scale service or a re grappling with more immediate problems like cache invalidation.\n\n### Speaker bio\n\nRegunath works at Flipkart where he is Principal Architect for Commerce and Supply Chain platforms. He also leads Flipkart's open sou rce initiatives and is committer on a number of projects. Prior to Flipkar t\, he architected and built Aadhaar - the world's largest biometric ident ity platform. His area of primary interest is large scale distributed syst ems. \nMore about him:\n\nhttps://github.com/regunathb/\n\nhttps://twitter .com/RegunathB\n GEO:12.943163506799058;77.596340412274 LAST-MODIFIED:20230810T072606Z LOCATION:Audi 2 - NIMHANS Convention center\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2015/schedule/building-tiered-data-s tores-using-aesop-to-bridge-sql-and-nosql-systems-LmE2pAiPNSHjY7gTpN7P8R BEGIN:VALARM ACTION:display DESCRIPTION:Building tiered data stores using Aesop to bridge SQL and NoSQ L systems in Audi 2 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Visualising Multi Dimensional Data DTSTART:20150717T093500Z DTEND:20150717T102000Z DTSTAMP:20260423T114800Z UID:session/2oKPB2ZAgorAhuPbTzEYZm@hasgeek.com SEQUENCE:2 CATEGORIES:Full Talk,Intermediate CREATED:20150706T050218Z DESCRIPTION:Even though exploring data visually is an integral part of the data analytic pipeline\, we struggle to visually explore data once the nu mber of dimensions go beyond three. This talk will focus on showcasing tec hniques to visually explore multi dimensional data (p > 3). The aim would be show examples of each of following techniques\, potentially using one e xemplar dataset.\n\n**Standard 2D/3D Approaches**\n\n- Aesthetics e.g. Col or\, Size\, Shape\n- Small Multiples e.g. Trellis / Facets\n- Matrices Vie ws e.g. SPLOMs\n- 3D Scatterplot\n\n**Geometric Transformation Approaches* *\n\n- Alternate Coordinates e.g. Parallel\, Star\n- Projections e.g. Dime nsionality Reduction\n- Tablelens\n\n**Glyph based Approaches**\n\n- Star glyphs\n- Stick Figures\n\n**Pixel based Approaches**\n\n- Pixel bar chart s\n- Space filling curves\n\n**Stacked based Approaches**\n\n- Dimensional Stacking\n- Hierarchical Axis\n- Treemaps\n\nThe talk will also explore t he role of interaction approaches to enhance our ability to visually explo re the multi dimensional data.\n\n**Interactive Approaches**\n\n- Navigati on - Pan\, Zoom\, Scale\, Rotate\n- Selection & Annotation\n- Filtering - Highlighting\, Brushing and Linking\n- Layering\n- Dynamic Queries\n\n### Speaker bio\n\nAmit Kapoor is interested in learning and teaching the craf t of telling visual stories with data. He uses storytelling and data visua lization as tools for improving communication\, persuasion and leadership. He conducts workshops and trainings for corporates\, non-profits\, colleg es\, and individuals at [narrativeVIZ Consulting](http://narrativeviz.com) . He also teaches storytelling with data as an invited guest faculty in ex ecutive courses at IIM Bangalore and IIM Ahmedabad and design courses at N ID\, Bangalore.\n\nHis background is in strategy consulting in using data- driven stories to drive change across organizations and businesses. He has more than 12 years of management consulting experience\, first with AT Ke arney in India\, then with Booz & Company in Europe and more recently for startups in Bangalore. He did his B.Tech in Mechanical Engineering from II T\, Delhi and PGDM (MBA) from IIM\, Ahmedabad. You can find more about him at [amitkaps.com](http://amitkaps.com) and tweet him at [@amitkaps](http: //twitter.com/amitkaps)\n GEO:12.943163506799058;77.596340412274 LAST-MODIFIED:20230810T072606Z LOCATION:Audi 1 - NIMHANS Convention center\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2015/schedule/visualising-multi-dime nsional-data-2oKPB2ZAgorAhuPbTzEYZm BEGIN:VALARM ACTION:display DESCRIPTION:Visualising Multi Dimensional Data in Audi 1 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Deep learning BOF DTSTART:20150717T093500Z DTEND:20150717T104000Z DTSTAMP:20260423T114800Z UID:session/JPzmAU9JeeJQfAb5CDmzBW@hasgeek.com SEQUENCE:0 CREATED:20150715T074057Z DESCRIPTION:\n GEO:12.943163506799058;77.596340412274 LAST-MODIFIED:20150715T083607Z LOCATION:BoF space - NIMHANS Convention center\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com BEGIN:VALARM ACTION:display DESCRIPTION:Deep learning BOF in BoF space in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Building Recommender system DTSTART:20150717T102000Z DTEND:20150717T104000Z DTSTAMP:20260423T114800Z UID:session/Rp8zRt2j9DGYWJb2nqXHtt@hasgeek.com SEQUENCE:2 CATEGORIES:Crisp Talk,Beginner CREATED:20150705T134414Z DESCRIPTION:Recommender Systems solves matrix sparsity problem. And this i dea of predicting sparse values can be applied for various problems across domains. I have used recommender systems to identify audience clusters fo r a conference\, recommending new jokes to users based on the past jokes t hey liked\, and few kaggle problems. \n\nOne of the key events that ener gized research in recommender systems was the Netflix prize. Netflix spons ored a competition\, that could take an offered dataset of over 100 millio n movie ratings and return recommendations that were 10% more accurate tha n those offered by the company's existing recommender system.\n\nRecommend er systems typically produce a list of recommendations in one of the two w ays - through collaborative or content-based filtering. Would like to cove r both of them with the implementation and mathematics involved.\n\n### Sp eaker bio\n\nCurrently part of data science team at Fidelity Investments\, Business Analytics and Research. \n\nMaster's in Mathematics from BITS\, Pilani.\nAm an open-source enthusiast and Kaggler. \n\nhttps://in.linkedin .com/in/kswaroop\n GEO:12.943163506799058;77.596340412274 LAST-MODIFIED:20230810T072606Z LOCATION:Audi 2 - NIMHANS Convention center\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2015/schedule/building-recommender-s ystem-Rp8zRt2j9DGYWJb2nqXHtt BEGIN:VALARM ACTION:display DESCRIPTION:Building Recommender system in Audi 2 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Escher - democratizing beautiful visualizations DTSTART:20150717T102000Z DTEND:20150717T104000Z DTSTAMP:20260423T114800Z UID:session/Ce6gP66wqFpPCcVLrDh9FX@hasgeek.com SEQUENCE:2 CATEGORIES:Crisp Talk,Beginner CREATED:20150706T050227Z DESCRIPTION:Escher.jl is a tool that lets you create and deploy web-based user interfaces flush with plots of data\, infographics\, rich-text\, math ematical typesetting\, and TeX-style layouts. All of which can update dyna mically as the data changes. The user only needs to work with one language - Julia. The overhead of writing HTML\, CSS and JavaScript disappears\, a nd the web is accessible to the average data scientist. What's more\, the interfaces made with Escher look great by default.\n\n### Speaker bio\n\nI am Shashi\, I can be found at https://github.com/shashi and https://twitt er.com/g0wda. I build user interface tools for the [Julia](http://julialan g.org) community. Of late\, I have been working on Escher\, on which my ta lk is going to be about.\n GEO:12.943163506799058;77.596340412274 LAST-MODIFIED:20230810T072606Z LOCATION:Audi 1 - NIMHANS Convention center\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2015/schedule/escher-democratizing-b eautiful-visualizations-Ce6gP66wqFpPCcVLrDh9FX BEGIN:VALARM ACTION:display DESCRIPTION:Escher - democratizing beautiful visualizations in Audi 1 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Evening tea break DTSTART:20150717T104000Z DTEND:20150717T111500Z DTSTAMP:20260423T114800Z UID:session/C5WfZkwj3f3MGF7Leqy9xf@hasgeek.com SEQUENCE:0 CREATED:20150705T133813Z DESCRIPTION:\n LAST-MODIFIED:20150707T074025Z LOCATION:Bangalore ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com BEGIN:VALARM ACTION:display DESCRIPTION:Evening tea break in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Two Years Wiser: The Nilenso Experiment DTSTART:20150717T111500Z DTEND:20150717T121500Z DTSTAMP:20260423T114800Z UID:session/H4e698ceDrs9AdXC2vuNPw@hasgeek.com SEQUENCE:2 CATEGORIES:Full Talk,Beginner CREATED:20150705T133901Z DESCRIPTION:When we describe nilenso as an "employee-owned democracy"\, pe ople are immediatley interested. "What does that even mean?" is a common f irst response. We'll usually respond by describing "employee-owned" in leg al terms and once people understand our structure\, they often have a diff icult time imagining that we can ever get anything accomplished: how are d ecisions are made democratically? who is responsible for outcomes? who has to have the uncomfortable conversations? what processes do we follow?\n\n An external observer often mistakes nilenso for a hippie commune or a disa sterous management case study waiting to happen. However\, when we retrosp ect over the past two years of our existence\, we see that the more ration ally and scientifically we approach any given situation\, the better the o utcome. We have certainly had our share of failures\, but open communicati on and transparency allows us to learn from them. Planning and projection is always uncertain\, but if we analyze the data and are honest with ourse lves\, the future is never quite as scary as it seems to our imaginations initially.\n\nAs we walk through two years of co-operative life\, you'll h ear about our genesis\, our ugliest client\, our hardest conversations\, o ur biggest successes\, our biggest mistakes\, and all that we have left to accomplish. Underpinning all of this are our purpose\, our goals\, and ou r ethos... themselves forever in flux as we learn.\n\n### Speaker bio\n\nS teven has worked for international consultancies\, early-stage startups\,\ nslow-saas-ramp-of-death startups\, trading firms\, the Canadian governmen t\,\nintelligence agencies\, and a little employee-owned tech coop here in Bangalore.\n\nAcross these organizations\, he’s seen a lot of project m anagement and team\nmanagement strategies. Some work by flow. Some work by force. Some never work. It’s\nbecome a mission in recent years to figur e out not just what works for him\, but\nwhat general toolset our companie s can use the world over to create more meaningful\,\nenjoyable places to work.\n GEO:12.943163506799058;77.596340412274 LAST-MODIFIED:20230810T072606Z LOCATION:Audi 1 - NIMHANS Convention center\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2015/schedule/two-years-wiser-the-ni lenso-experiment-H4e698ceDrs9AdXC2vuNPw BEGIN:VALARM ACTION:display DESCRIPTION:Two Years Wiser: The Nilenso Experiment in Audi 1 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Processing large data with Apache Spark DTSTART:20150717T121500Z DTEND:20150717T130000Z DTSTAMP:20260423T114800Z UID:session/J9LQ6zwW2D4bu3JCLuGcp4@hasgeek.com SEQUENCE:2 CATEGORIES:Full Talk,Intermediate CREATED:20150705T133848Z DESCRIPTION:In this session\, Ravi will cover Apache Spark overview with i ts unique features using in large data systems. He will get more details i nto Spark EcoSystem\,Architecture\, Elements and comparison with MapReduce . He will also touch up on its languages support with working demo session .\n\n### Speaker bio\n\nRavi is in in IT industry for 11+ years. Ravi work s for Cisco as Technical Leader and part of Cisco service team .He complet ed MS from BITS and BE from University of Madras. He has well experience in building highly distributable systems using multi-tier architecture. Hi s interest on exploring new technologies and tools.\n\nhttps://www.linkedi n.com/in/venkatanagaravi\n GEO:12.943163506799058;77.596340412274 LAST-MODIFIED:20230810T072606Z LOCATION:Audi 2 - NIMHANS Convention center\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2015/schedule/processing-large-data- with-apache-spark-J9LQ6zwW2D4bu3JCLuGcp4 BEGIN:VALARM ACTION:display DESCRIPTION:Processing large data with Apache Spark in Audi 2 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Recommendation System beyond traditional Collaborative filtering DTSTART:20150717T121500Z DTEND:20150717T130000Z DTSTAMP:20260423T114800Z UID:session/U24k4HkMm1FzyZ9N1DyLx3@hasgeek.com SEQUENCE:2 CATEGORIES:Full Talk,Intermediate CREATED:20150705T134442Z DESCRIPTION:Though Collaborative filtering works quite well for companies like NetFlix but here in Snapdeal we are catering 12M huge product catalog and more than 100 categories which again comprised of 20-30 subcategories each. For us only Collaborative filtering doesn't work well\, because of the wide catalog and implicit feedback capturing instead of explicit and h ence we developed a recommendation system which considers various other fa ctors beyond collabarative Filtering.\n\nIn this session I would be discus sing other factors (mentioned below) and their mathematical models that we have considered while building custom recommendation system for generatin g more personalized and relevant recommendations.\n\n1. User Category Affi nity (to some more granular level)\n2. Content based product similarity\n3 . product which goes well with already bought products.\n4. predicting the repurchase of already purchased products.\n5. Suggesting trending product s based on user's affinity.\n6. Capturing user's feedback (implicit) to ou r served recommendations and use to improve relevancy.\n7. Collaborative f iltering (we have also used this but with some weight-age)\n\nFinally I wo uld be concluding session with technical challenges in building scalable r ecommendation system with massive datasets and serving these recommendatio ns in realtime.\n\n### Speaker bio\n\nGagan Deep Juneja is a Lead Engineer at Snapdeal and is leading several initiatives related to user personaliz ation. He has close to 7 years of experience in the Software Industry. He has worked on several projects using Java/J2ee and Hadoop as the primary t echnology. He has been working with Big data technologies such as hadoop\, spark\, cascading\, pig\, hive\, blur for the past couple of years. He ha s great interest in Machine learning concepts and working with supervised and unsupervised algorithms to get value out of data. He has an inclinatio n to open source technologies and likes to delve into new frameworks. He i s a committer and PPMC member to Apache Blur (incubating). He has spoken a t various meetup groups in the past. He is a active blogger\, and in his l eisure time loves exploring new technologies and keeping himself updated w ith latest trends.\n GEO:12.943163506799058;77.596340412274 LAST-MODIFIED:20230810T072606Z LOCATION:Audi 1 - NIMHANS Convention center\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2015/schedule/recommendation-system- beyond-traditional-collaborative-filtering-U24k4HkMm1FzyZ9N1DyLx3 BEGIN:VALARM ACTION:display DESCRIPTION:Recommendation System beyond traditional Collaborative filteri ng in Audi 1 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Registrations and check-ins DTSTART:20150718T033000Z DTEND:20150718T040000Z DTSTAMP:20260423T114800Z UID:session/BdGjcuwA1UXPYYktc15noq@hasgeek.com SEQUENCE:0 CREATED:20150629T112629Z DESCRIPTION:\n LAST-MODIFIED:20150707T074553Z LOCATION:Bangalore ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com BEGIN:VALARM ACTION:display DESCRIPTION:Registrations and check-ins in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Introduction to Deep Learning DTSTART:20150718T040000Z DTEND:20150718T053000Z DTSTAMP:20260423T114800Z UID:session/K7HjAgKiqKZRAL4W6ay8hf@hasgeek.com SEQUENCE:2 CATEGORIES:Workshop,Intermediate CREATED:20150629T113933Z DESCRIPTION:* What is deep learning?\n* *Motivation:* Some use cases where it has produced state-of-art results\n* Supervised learning (multi-layer perceptron\, deep convolution networks)\n* Unsupervised learning (autoenco ders) *time permitting* \n* Overview of a few libraries and the impact of GPUs (Some practical thoughts on hardware and software)\n* Hands-on modeli ng a simple classification problem\n\nWe would be using Python-based libra ries to do the hands-on. \n\nThe data and software requirements would be p osted onto the github repository one week prior to the workshop. \n\nThe r epository for this workshop:\nhttps://github.com/rouseguy/intro2deeplearni ng/\n\n### Speaker bio\n\n* Bargava Subramanian is a Senior Statistician(D ata Scientist) at Cisco Systems\, India. He has a Masters from University of Maryland\, College Park\, USA.\n\n* Raghotham is a full-stack developer at RedMart. He has a Masters from BITS\, Pilani. In his previous role\, h e was instrumental in architecting an analytics platform for a wearable de vices company.\n GEO:12.943163506799058;77.596340412274 LAST-MODIFIED:20230810T072606Z LOCATION:Audi 3 - NIMHANS Convention center\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2015/schedule/introduction-to-deep-l earning-K7HjAgKiqKZRAL4W6ay8hf BEGIN:VALARM ACTION:display DESCRIPTION:Introduction to Deep Learning in Audi 3 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Understanding supervised machine learning hands on! DTSTART:20150718T040000Z DTEND:20150718T053000Z DTSTAMP:20260423T114800Z UID:session/LWrNLeic6iWaFv293DNFAG@hasgeek.com SEQUENCE:2 CATEGORIES:Workshop,Beginner CREATED:20150707T082834Z DESCRIPTION:### What will participants gain?\n[Here's a mind map showing t he overall picture of what will be covered in the workshop.](http://www.sl ideshare.net/HarshadSaykhedkar/ml-workshop-jul2015mm)\n\n* Design of large software systems is a study and practice of making trade-offs (e.g. CAP t heorem\, time Vs. space complexities\, time to build Vs. maintainability). Same is true for machine learning applications. This workshop will help y ou to clearly understand what those trade-offs are and how to make one.\n* Black box way of building ML applications (use X because company F/G/H us es it) can only get us to a point. Workshop will instead help you understa nd all core ideas of ML in clear intuitive fashion.\n* There are multiple problems in a ML application : modelling\, information representation\, na ture of costs etc. The workshop will give you the big picture and practica l advice on tackling the problems.\n* Understand and apply some/all the fo llowing models\,\n * simple neighbourhood based models\n * regressio n models\n * decision trees / random forests / ensemble methods\n * support vector machines\n * neural networks / (if time permits) basics of deep learning\n* Gain sound understanding of\n * training\, testing\ , cross validation\, evaluation.\n * feature engineering practices for various domains.\n * how to debug models and decide next steps.\n \n### Workshop schedule / plan?\nThis will be a 4 hour workshop with a short break in the middle. The broad outline is as follows\,\n\n* I ntroduction : 10 minutes\n* Core ideas\, cost functions\, likelihoods\, op timizations and best fit : 20 minutes\n* Information representation\, simp le representations\, linear and generalized linear models : 30 minutes\n* Complex and non-linear representations\, feature engineering : 20 minutes\ n* More models : SVMs\, tree based models\, neural networks\, introduction to deep learning : 50 minutes\n* Domain understanding\, asymetric costs\, evaluation methods and metrics : 30 minutes\n* Trade-offs : model complex ity Vs. representation complexity\, interpretability\, cost of gathering d ata\, model selections : 30 minutes\n* Summary\, big picture\, question an d answers : 30 minutes\n \n### FAQ\n\n* How much machine learning s hould I know already?\n \n We expect you to know bare minimum basics l ike supervised Vs. unsupervised machine learning model. If you know what i s a linear regression\, it should be good enough.\n\n* I don't know Python . Is this workshop for me?\n\n Yes. As long as you know basics of progr amming and have written atleast some code in any language.\n\n* How much p rogramming should I know to attend?\n\n You should know basic programmi ng (loops\, conditional expressions\, variable assignments\, reading files \, performing some data manipuation on them).\n\n* What about the data and the code to be used at the time of workshop?\n\n We will using [this g ithub repository](https://github.com/harshadss/ml_workshop_jul15) to share code and data.\n\n* Why not cover unsupervised learning/semi-supervised l earning/some other fancy model X?\n\n We will focus on depth and try to cover few topics well.\n\n* Will the workshop cover Apache Spark/Hadoop/M ahout or X library/ecosystem?\n\n No. This is an ideas/algorithms talk and libraries will just serve as means for understanding. Different librar ies/ecosystems are likely to be covered in depth by other speakers.\n\n### Speaker bio\n\nHarshad leads the machine learning and data team @ Sokrati \, an advertising technology and analytics company based out of Pune. He h as spent 6 years in applying statistical models in variety of domains like insurance\, banking\, telecom and advertising. He has experience with man y tools in the data ecosystem like Python\, R\, Clojure\, Hadoop\, Spark e tc. He spends time learning theory and applications of machine learning mo dels from simple regression to deep learning. Harshad holds a master's deg ree from Indian Institute of Techonology\, Mumbai.\n GEO:12.943163506799058;77.596340412274 LAST-MODIFIED:20230810T072606Z LOCATION:Audi 2 - NIMHANS Convention center\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2015/schedule/understanding-supervis ed-machine-learning-hands-on-LWrNLeic6iWaFv293DNFAG BEGIN:VALARM ACTION:display DESCRIPTION:Understanding supervised machine learning hands on! in Audi 2 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Morning tea break DTSTART:20150718T053000Z DTEND:20150718T060000Z DTSTAMP:20260423T114800Z UID:session/HuXaFmQMEF9kJPL3mfgfKf@hasgeek.com SEQUENCE:0 CREATED:20150707T074459Z DESCRIPTION:\n LAST-MODIFIED:20150707T074600Z LOCATION:Bangalore ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com BEGIN:VALARM ACTION:display DESCRIPTION:Morning tea break in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Understanding supervised machine learning hands-on - continued DTSTART:20150718T060000Z DTEND:20150718T080000Z DTSTAMP:20260423T114800Z UID:session/B2HE27KjSpHXE1zSJkYKFv@hasgeek.com SEQUENCE:0 CREATED:20150707T083135Z DESCRIPTION:\n GEO:12.943163506799058;77.596340412274 LAST-MODIFIED:20150707T083154Z LOCATION:Audi 2 - NIMHANS Convention center\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com BEGIN:VALARM ACTION:display DESCRIPTION:Understanding supervised machine learning hands-on - continued in Audi 2 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Introduction to deep learning – continued DTSTART:20150718T060000Z DTEND:20150718T080000Z DTSTAMP:20260423T114800Z UID:session/TZUdYTJNkj7GsQnvaFgvUq@hasgeek.com SEQUENCE:0 CREATED:20150707T074538Z DESCRIPTION:\n GEO:12.943163506799058;77.596340412274 LAST-MODIFIED:20150707T083157Z LOCATION:Audi 3 - NIMHANS Convention center\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com BEGIN:VALARM ACTION:display DESCRIPTION:Introduction to deep learning – continued in Audi 3 in 5 min utes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Lunch DTSTART:20150718T080000Z DTEND:20150718T084500Z DTSTAMP:20260423T114800Z UID:session/RkkS2rQPtxrYfWQaVBfFZS@hasgeek.com SEQUENCE:0 CREATED:20150707T082519Z DESCRIPTION:\n LAST-MODIFIED:20150707T082859Z LOCATION:Bangalore ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com BEGIN:VALARM ACTION:display DESCRIPTION:Lunch in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Introduction to deep learning continued DTSTART:20150718T084500Z DTEND:20150718T103000Z DTSTAMP:20260423T114800Z UID:session/CdFbN7AfBtuQHNJPY7PEb1@hasgeek.com SEQUENCE:0 CREATED:20150707T082543Z DESCRIPTION:\n GEO:12.943163506799058;77.596340412274 LAST-MODIFIED:20150707T082856Z LOCATION:Audi 3 - NIMHANS Convention center\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com BEGIN:VALARM ACTION:display DESCRIPTION:Introduction to deep learning continued in Audi 3 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Igniting your data with Apache Spark DTSTART:20150718T084500Z DTEND:20150718T103000Z DTSTAMP:20260423T114800Z UID:session/Lvq38SRTWUefn9rA1wS4sQ@hasgeek.com SEQUENCE:2 CATEGORIES:Workshop,Beginner CREATED:20150707T083925Z DESCRIPTION:The workshop aims to help the audience understand the architec ture of Spark and work with it’s core API. The workshop also aims to hel p the audience understand how to build and test data applications based on Spark\n\n- Introduction to Spark and in memory computing\n- Setting up sp ark locally\n- Spark architecture\n- Working with Spark APIs\n\nSpark\, a step forward:\n\n- Spark Streaming\, Spark SQL\, MlLib\, Dataframes\n- Dep loying spark applications\n- Building spark applications\n- Spark gotchas and best practices\n\nSpark\, a deep dive (based on audience interest and if time permits)\n\n- Spark RDDs -> various RDDs that are part of Spark\n- Spark and other input sources i.e. Cassandra\, hive etc\n\n### Speaker bi o\n\nYagnik is a software developer at Shopify.\nlinkedin.com/in/yagnik\n GEO:12.943163506799058;77.596340412274 LAST-MODIFIED:20230810T072606Z LOCATION:Audi 2 - NIMHANS Convention center\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2015/schedule/igniting-your-data-wit h-apache-spark-Lvq38SRTWUefn9rA1wS4sQ BEGIN:VALARM ACTION:display DESCRIPTION:Igniting your data with Apache Spark in Audi 2 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Evening tea break DTSTART:20150718T103000Z DTEND:20150718T110000Z DTSTAMP:20260423T114800Z UID:session/7g5U5bFVCMqK1wDPxZsBMf@hasgeek.com SEQUENCE:0 CREATED:20150707T082605Z DESCRIPTION:\n LAST-MODIFIED:20150707T082915Z LOCATION:Bangalore ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com BEGIN:VALARM ACTION:display DESCRIPTION:Evening tea break in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Introduction to deep learning – continued DTSTART:20150718T110000Z DTEND:20150718T123000Z DTSTAMP:20260423T114800Z UID:session/24rWTLRPmcCtaVnyK8sUxG@hasgeek.com SEQUENCE:0 CREATED:20150707T082644Z DESCRIPTION:\n GEO:12.943163506799058;77.596340412274 LAST-MODIFIED:20150707T082917Z LOCATION:Audi 3 - NIMHANS Convention center\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com BEGIN:VALARM ACTION:display DESCRIPTION:Introduction to deep learning – continued in Audi 3 in 5 min utes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Igniting your data with Apache Spark - continued DTSTART:20150718T110000Z DTEND:20150718T130000Z DTSTAMP:20260423T114800Z UID:session/QGaZfgXr66Dny3XjosFNaQ@hasgeek.com SEQUENCE:0 CREATED:20150707T084010Z DESCRIPTION:\n GEO:12.943163506799058;77.596340412274 LAST-MODIFIED:20150707T084052Z LOCATION:Audi 2 - NIMHANS Convention center\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com BEGIN:VALARM ACTION:display DESCRIPTION:Igniting your data with Apache Spark - continued in Audi 2 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT END:VCALENDAR