BEGIN:VCALENDAR VERSION:2.0 PRODID:-//HasGeek//NONSGML Funnel//EN DESCRIPTION:On theory and concepts in Machine Learning\, Deep Learning and Artificial Intelligence. Formerly Deep Learning Conf. X-WR-CALDESC:On theory and concepts in Machine Learning\, Deep Learning an d Artificial Intelligence. Formerly Deep Learning Conf. NAME:Anthill Inside 2017 X-WR-CALNAME:Anthill Inside 2017 REFRESH-INTERVAL;VALUE=DURATION:PT12H SUMMARY:Anthill Inside 2017 TIMEZONE-ID:Asia/Kolkata X-PUBLISHED-TTL:PT12H X-WR-TIMEZONE:Asia/Kolkata BEGIN:VEVENT SUMMARY:Check-in DTSTART:20170729T033000Z DTEND:20170729T040000Z DTSTAMP:20260409T160148Z UID:session/WWe9m4j8FQArUhikceazau@hasgeek.com SEQUENCE:0 CREATED:20170717T102821Z DESCRIPTION:\n LAST-MODIFIED:20170720T123739Z LOCATION:Bangalore ORGANIZER;CN="Anthill Inside":MAILTO:no-reply@hasgeek.com BEGIN:VALARM ACTION:display DESCRIPTION:Check-in in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:AI in healthcare: problems we solve at SigTuple. DTSTART:20170729T040000Z DTEND:20170729T042000Z DTSTAMP:20260409T160148Z UID:session/Kb6RoHZYaM2N3FVPrqUKU4@hasgeek.com SEQUENCE:0 CREATED:20170616T072437Z DESCRIPTION:\n LAST-MODIFIED:20170722T170927Z LOCATION:Banquet Hall - MLR Whitefield\nIN ORGANIZER;CN="Anthill Inside":MAILTO:no-reply@hasgeek.com BEGIN:VALARM ACTION:display DESCRIPTION:AI in healthcare: problems we solve at SigTuple. in Banquet Ha ll in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Deep Reinforcement Learning: a tutorial. DTSTART:20170729T040000Z DTEND:20170729T044500Z DTSTAMP:20260409T160148Z UID:session/Nwrs2NcMEeM9EGLEjw1f4D@hasgeek.com SEQUENCE:0 CREATED:20170721T195413Z DESCRIPTION:\n LAST-MODIFIED:20170722T170458Z LOCATION:Auditorium - MLR Whitefield\nIN ORGANIZER;CN="Anthill Inside":MAILTO:no-reply@hasgeek.com BEGIN:VALARM ACTION:display DESCRIPTION:Deep Reinforcement Learning: a tutorial. in Auditorium in 5 m inutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Unsupervised and semi-supervised Deep Learning for medical imaging . DTSTART:20170729T042000Z DTEND:20170729T050500Z DTSTAMP:20260409T160148Z UID:session/FLDd9heQ79Bw1W6Ex6iSAN@hasgeek.com SEQUENCE:2 CATEGORIES:Full talk,Advanced CREATED:20170722T111733Z DESCRIPTION:### Introduction - [10 mins]\n* Deep learning in Medical Imagi ng\n* Diagnosing Glioblastoma in brain with MRI\n* Annotation problem\n* P re-processing\n### Auto-encoders - [15 mins]\n* Auto-encoders\n* Unsupervi sed learning\n* Pre-training deep autoencoders on unlabelled data\n* Fine- tuning on limited labelled data\n### Unsupervised learning - [5 mins]\n* N ovelty detection using autoencoder\n* Segmentation using unsupervised lear ning\n### Results - [5 mins]\n* Post-processing\n* Segmentation results w. r.t state-of-the-art\n### Conclusions - [5 mins]\n* Unsupervised + Supervi sed in one go\n* Ladder Networks\n* Y-Nets\n* The future of unsupervised l earning\n\n### Speaker bio\n\nKiran Vaidhya holds a dual degree (B.Tech + M.Tech) in Engineering Design (specialization in Biomedical Design) from I IT Madras. He has been heavily involved in Computer Vision and Medical Ima ging for the past 4 years. His Master's thesis was on brain tumor segmenta tion from MRI using Semi-Supervised Deep Learning. His work has been publi shed and accepted by leading medical imaging journals like [MICCAI](http:/ /www.miccai.org/).\n\nPost his graduation\, he joined [Predible Health](ht tp://prediblehealth.com/) and is currently working as an Algorithms Resear cher for CAD (Computer Aided Diagnosis) system design. Deep learning is a natural part of his work in order to derive data-driven insights. He has b een actively involved in the development of Torch and has extensive experi ence with Theano and TensorFlow.\n LAST-MODIFIED:20230810T072606Z LOCATION:Banquet Hall - MLR Whitefield\nIN ORGANIZER;CN="Anthill Inside":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/anthillinside/2017/schedule/unsupervised-and-semi- supervised-deep-learning-for-medical-imaging-FLDd9heQ79Bw1W6Ex6iSAN BEGIN:VALARM ACTION:display DESCRIPTION:Unsupervised and semi-supervised Deep Learning for medical ima ging. in Banquet Hall in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Hitchhiker’s guide to Generative Adversarial Networks (GANs). DTSTART:20170729T044500Z DTEND:20170729T053000Z DTSTAMP:20260409T160148Z UID:session/TFLR3dXb6kUsbHQUVU3mJy@hasgeek.com SEQUENCE:2 CATEGORIES:Full talk,Intermediate CREATED:20170717T103623Z DESCRIPTION:### Motivation + Overview of Generative Networks [5 mins]\n\n- Reasoning for interest in generative networks\n- Overview of options for such networks\n- Issues and motivations for developing GANs\n\n--- \n\n### GANs - Fundamentals [12 mins]\n\nIntroduce the framework under which GANs can be studied. The following ideas will be discussed to enable a clear u nderstanding of the fundaments\n\n- GANs as games played betweene adversar ial networks\n- Generator v. Discriminator networks and "goals" for each n etwork\n- Loss function options and training process for GANs\n- DCGAN arc hitecture\n\n--- \n\n### GANs - Applications and Recent Developments [10 m ins]\n\nAn overview of developments in the past year. Touching upon each o f the following\, covering how they were developed\, salient contributions and applications that can be realized.\n\n- SRGANs for single image super -resolution\n- Interactive GANs for image generation\n- pix2pix\, image to image transalation\n- Conditional GANs for text to image synthesis\n- Ima ge inpainting using DCGANs\n- Using multiple GANs for cross-domain applica tion and "style transfer"\n\n--- \n\n### Issues & Improvements [8 mins]\n\ nOverview of the issues affecting GANs and the steps that are being taken to combat them. \n\n- Recent developments that are helping solve some of t he known stability problems\n- Tricks to optimize development and training of GANs\n- Looking at impact of GANs going forward and effect on other as pects of generative modelling\n\n### Speaker bio\n\nI am a member of the d ata science team at [Semantics3](https://www.semantics3.com) - building da ta-powered software for ecommerce-focused companies. Over the years\, I ha ve had the chance to dabble in various fields covering data processing\, p ipeline setup\, database management and data science. When not picking loc ks\, or scuba diving\, I usually blog about my technical adventures at our [team’s engineering blog](https://engineering.semantics3.com/@ramananb) .\n LAST-MODIFIED:20230810T072606Z LOCATION:Auditorium - MLR Whitefield\nIN ORGANIZER;CN="Anthill Inside":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/anthillinside/2017/schedule/hitchhikers-guide-to-g enerative-adversarial-networks-gans-TFLR3dXb6kUsbHQUVU3mJy BEGIN:VALARM ACTION:display DESCRIPTION:Hitchhiker’s guide to Generative Adversarial Networks (GANs) . in Auditorium in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Identifying urban makeshift communities using satellite imagery an d geo-coded data. DTSTART:20170729T050500Z DTEND:20170729T052500Z DTSTAMP:20260409T160148Z UID:session/SJehg4JTWcvq2r8hmVoZKk@hasgeek.com SEQUENCE:2 CATEGORIES:Crisp talk,Intermediate CREATED:20170722T114211Z DESCRIPTION:This session will be organized as:\n\n- Problem Description\n- Data Gathering \n - Challenges with data\n- Various Modeling Approache s\n - Satellite Imaging Approach\n - Data Sources\n - Com puter Vision Approach (No labeled data)\n - Supervised Learning App roach\n - Transfer Learning Approach\n - Geo-Coded Data based Ap proach\n - Data Sources\n - Undersampling Approach\n - Stratified sampling Approach\n- Closing Thoughts/Learning\n\n### Speaker bio\n\nI volunteer with DataKind Bangalore. Professionally\, I am working at Infrrd AI on Machine Learning.\n LAST-MODIFIED:20230810T072606Z LOCATION:Banquet Hall - MLR Whitefield\nIN ORGANIZER;CN="Anthill Inside":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/anthillinside/2017/schedule/identifying-urban-make shift-communities-using-satellite-imagery-and-geo-coded-data-SJehg4JTWcvq2 r8hmVoZKk BEGIN:VALARM ACTION:display DESCRIPTION:Identifying urban makeshift communities using satellite imager y and geo-coded data. in Banquet Hall in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Morning beverage break DTSTART:20170729T053000Z DTEND:20170729T060000Z DTSTAMP:20260409T160148Z UID:session/AW9vCRAJ5fbdeMrkAsDjyL@hasgeek.com SEQUENCE:0 CREATED:20170616T072048Z DESCRIPTION:\n LAST-MODIFIED:20170722T103009Z LOCATION:Bangalore ORGANIZER;CN="Anthill Inside":MAILTO:no-reply@hasgeek.com BEGIN:VALARM ACTION:display DESCRIPTION:Morning beverage break in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Sponsored keynote: "AI: unleashing the next wave." DTSTART:20170729T060000Z DTEND:20170729T063000Z DTSTAMP:20260409T160148Z UID:session/MVT7S6azdrZVQ7VABdpxM8@hasgeek.com SEQUENCE:2 CATEGORIES:Full talk,Intermediate CREATED:20170718T040923Z DESCRIPTION:This talk provides with a high-level overview of Intel’s Ar tificial Intelligence (AI) vision and product portfolios. This talks star ts with where Intel sees opportunity in various verticals and industries f or AI and we will take one example of Intel’s comprehensive AI strategy in action. This talk gives overview on both hardware\, software portfolio and also our developer outreach programs and engagements and a good primer for detailed technical talk later in the day by Intel Corporation\n\n### Speaker bio\n\nMilind Hanchinmani: Director\, Data center and IOT Enabling - APJ\, Intel Corporation\n\nMilind is currently the director of Develope r Relations Division @ Intel focusing on IOT and Data center. In this rol e he focuses on developing the software solutions market across APJ. Milin d drives engineering\, developer and partner marketing with ISV’s\, Acad emia and software developers in the region across data center\, IOT. \nMil ind joined Intel in 2001 and was part of Intel onsite team at Microsoft\, Redmond helping in improvements in the quality and performance of three ge nerations of the Microsoft® .NET Framework for Intel Architecture. Before that he was part of onsite team Microsoft\, Redmond in validating compile rs and performance improvements for Intel® Architecture. He has great exp ertise about performance\, scalability\, benchmarking\nMilind has degree B E in Computer Science & Engg from SDM College of Engineering\, Dharwad.\n LAST-MODIFIED:20230810T072606Z LOCATION:Auditorium - MLR Whitefield\nIN ORGANIZER;CN="Anthill Inside":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/anthillinside/2017/schedule/ai-unleashing-the-next -wave-MVT7S6azdrZVQ7VABdpxM8 BEGIN:VALARM ACTION:display DESCRIPTION:Sponsored keynote: "AI: unleashing the next wave." in Auditori um in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Panel discussion: "When – and when not – to use Deep Learning\ , Machine Learning and AI?" DTSTART:20170729T063000Z DTEND:20170729T073000Z DTSTAMP:20260409T160148Z UID:session/9ZV9RZJGYkkYaSpHfrMWMu@hasgeek.com SEQUENCE:0 CREATED:20170616T071349Z DESCRIPTION:\n LAST-MODIFIED:20170729T024636Z LOCATION:Auditorium - MLR Whitefield\nIN ORGANIZER;CN="Anthill Inside":MAILTO:no-reply@hasgeek.com BEGIN:VALARM ACTION:display DESCRIPTION:Panel discussion: "When – and when not – to use Deep Learn ing\, Machine Learning and AI?" in Auditorium in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Lunch DTSTART:20170729T073000Z DTEND:20170729T083000Z DTSTAMP:20260409T160148Z UID:session/PP6MRHL7G189iZqjvxTznj@hasgeek.com SEQUENCE:0 CREATED:20170616T072403Z DESCRIPTION:\n LAST-MODIFIED:20170718T041127Z LOCATION:Bangalore ORGANIZER;CN="Anthill Inside":MAILTO:no-reply@hasgeek.com BEGIN:VALARM ACTION:display DESCRIPTION:Lunch in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Adversarial attacks on Deep Learning models. DTSTART:20170729T083000Z DTEND:20170729T091500Z DTSTAMP:20260409T160148Z UID:session/8KTDfmhDmC7yQZVYK6xCVY@hasgeek.com SEQUENCE:2 CATEGORIES:Full talk,Intermediate CREATED:20170722T132512Z DESCRIPTION:Introduction (~10 mins): I will talk about what adversarial im ages are\, talk about why they are a serious issue and convince that it an important and worthy excercise to study them.\nAdversarial attacks (~15 m ins): I will introduce some of the seminal works that revealed this intrig uing property of current deep learning models.\nTechniques to defend (10 m ins): This part of the talk is dedicated to discuss some of the most effec tive defending techniques to date against these attacks.\nInteraction (~ 5 mins): Q&A\n\n### Speaker bio\n\nKonda Reddy Mopuri is currently a PhD stu dent at the Department of Computational and Data Sciences\, Indian Institu te of Science (IISc)\, Bangalore. He obtained a masters degree from Indian Institute of Technology Kharagpur in Visual information processing and Em bedded systems. Before he joined IISc for a PhD\, he worked in Samsung Ind ia for a brief one year. His research interest is to apply deep learning t echniques to solve various computer vision problems. Lately\, he has been working towards learning and understanding deep learned visual representa tions. (weblink: https://sites.google.com/site/kreddymopuri/)\n LAST-MODIFIED:20230810T072606Z LOCATION:Banquet Hall - MLR Whitefield\nIN ORGANIZER;CN="Anthill Inside":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/anthillinside/2017/schedule/adversarial-attacks-on -deep-learning-models-8KTDfmhDmC7yQZVYK6xCVY BEGIN:VALARM ACTION:display DESCRIPTION:Adversarial attacks on Deep Learning models. in Banquet Hall i n 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Panel discussion: "AI and product." DTSTART:20170729T083000Z DTEND:20170729T093000Z DTSTAMP:20260409T160148Z UID:session/XyZEGgroGoqd35eVTG3nDi@hasgeek.com SEQUENCE:2 CATEGORIES:Birds of a Feather (BOF) session,Beginner CREATED:20170717T105011Z DESCRIPTION:Part 1: busting the AI myth\n\nCritique questions:\nDo you rea lly apply deep learning (or AI)\nWhat is the size of dataset that you play with\nWhat is more painful: collecting data or applying AI\nHow do you ma p effect/outcome of deep learning (or AI) on product metrics\n\nPart 2: da rk data\n\nThe data advantage\nThe user advantage\nBuilding a product that has a loop between users and data\nIs data your core advantage (and not t echnology)\nHow fast can your competitors have the same data (or technolog y)\n\nPart 3: building products\n\nB2B (or SAAS) product vs consulting\nSi nce you don't own data\, how do you build a scalable product\nIf there is too much customization required to serve every customer\, how do you build a business\n\nB2C product vs technology\nHow do you build defensibility w ith AI\nDo you build good to have features for existing market or a scalab le product to serve new market\nPatents vs 10X improvement in technology v ia research\n\nPart 4: product-customer experiences\n\nHas AI delivered on its promise so far\nWhat more needs to be done\n\n### Speaker bio\n\nThis panel discussion will be facilitated by Vijay Gabale. Vijay is co-founder and CTO of Huew\, an AI-powered Commerce Platform. Huew has been building a fashion commerce platform to provide exceptional shopping experiences t o the Internet consumers. The company has made several innovations in deep learning to process rich multi-media data (text\, image\, videos) to impr ove discovery\, search and personalization experiences of online consumers .\n LAST-MODIFIED:20230810T072606Z LOCATION:Auditorium - MLR Whitefield\nIN ORGANIZER;CN="Anthill Inside":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/anthillinside/2017/schedule/panel-on-ai-and-produc t-XyZEGgroGoqd35eVTG3nDi BEGIN:VALARM ACTION:display DESCRIPTION:Panel discussion: "AI and product." in Auditorium in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Keep calm and trust your model: on explainability of ML models. DTSTART:20170729T091500Z DTEND:20170729T100000Z DTSTAMP:20260409T160148Z UID:session/5JDRd7a5GYQU5Q62tkvaxr@hasgeek.com SEQUENCE:2 CATEGORIES:Full talk,Intermediate CREATED:20170722T131812Z DESCRIPTION:##### Motivation + Intro on Explainability (5 mins)\n- The nee d for explainability\n- Why are certain models not explainable? \n- Linear \, monotonic vs Non-linear\, non-monotonic functions\n\n##### Model Specif ic approaches to Explainability (15 mins)\nSpecial model specific methods\ , deep dive into a few of them :\n\n- Tree Interpreter for explaining Tree based models like Random Forest and Gradient Boosted Trees\n- Bayesian Ru le Lists\n- Model Specific Visualisations\n- Attention mechanism used to e xplain predictions\n- Generating explanations as a part of the model itsel f (cutting edge deep learning models from MIT and Berkeley that give an ex planation as additional output along with the predicted class/value)\n\n## ### Model Agnostic approaches to Explainability (10 mins)\n- Global scoped Surrogate models\, statistical interpretation tools like Variable Importa nce\, Residual plot etc\n- Local Interpretable Model-agnostic Explanations : recent research which works on any black box model\n- Layerwise Relevan ce Propagation for understanding Deep Learning\n - for CNNs\n - for RNNs (the source code for this was released just 15 days back! We'll be do ing a live demo of this method)\n\n##### Live Demo of the above approaches and Conclusion (10 mins)\n- Use open source tools in Python and learn how to make use of them to explain machine learning model predictions\n- Conc lusion with practical demonstrations and call to action to try out the too ls\n\n### Speaker bio\n\nCurrently working as a as Machine Learning Engine er at [datalog.ai](https://datalog.ai)\, working remotely from Kochi\, I'm entirely self-taught in the field\, and originally did Bachelors in Mecha nical Engineering from CUSAT. \nI have completed consulting projects in ML and AI with multiple startups and companies. \n\nPreviously I was a Techn ology Innovation Fellow with Kerala Startup Mission where I started a non- profit student community TinkerHub\, that has a focus on creating communit y spaces across colleges for learning the latest technologies.\n\nMy work on CNNs was the winning solution for IBM's Cognitive Cup challenge in 2016 and gave a talk on the same at the Super Computing conference SC16 at Sal t Lake City\, Utah : [Slides](https://speakerdeck.com/psbots/artnet-ibm-op enpower-cognitive-cup-contest-winning-talk)\n\nExplainability and Interpre tability of ML is one of my focus areas\, after having interacted with man y Business owners asking for the reasons behind the working of the predict ion models built for them.\n LAST-MODIFIED:20230810T072606Z LOCATION:Banquet Hall - MLR Whitefield\nIN ORGANIZER;CN="Anthill Inside":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/anthillinside/2017/schedule/keep-calm-and-trust-yo ur-model-on-explainability-of-machine-learning-models-5JDRd7a5GYQU5Q62tkva xr BEGIN:VALARM ACTION:display DESCRIPTION:Keep calm and trust your model: on explainability of ML models . in Banquet Hall in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Synthetic gradients: decoupling layers of neural nets. DTSTART:20170729T093000Z DTEND:20170729T101500Z DTSTAMP:20260409T160148Z UID:session/75WsziF58WKdqrJ3VKkSuN@hasgeek.com SEQUENCE:2 CATEGORIES:Full talk,Intermediate CREATED:20170718T041109Z DESCRIPTION:### Refresher on Back propagation [5 mins]\n * Basics\n\n### Problems with Back propagation [5 mins]\n * Forward locking\n * Backward locking\n * Update locking\n * Impact of locking\n\n### Why does it mat ter [1 mins]\n\n### Applications [3 mins]\n\n### Solution [12 mins]\n * S ynthetic Gradients\n * Breaking backward & update locking\n \n### Result s [5 mins]\n * Backprop vs Synthetic Gradients\n\n### Complete unlock [2 mins]\n * Breaking forward locking\n\n### Closing remarks [3 mins]\n\nTo facilitate better understanding\, I will be giving a github repo as a take away so that the audience can go back\, download the code and play with i t.\nCode assosiated with this talk : https://github.com/anujgupta82/Synthe tic_Gradients\n\n### Speaker bio\n\n__Anuj Gupta__ is a senior ML research er at Freshdesk\; working in the area NLP\, Machine Learning\, Deep learni ng. Earlier he was heading ML efforts at Airwoot(Now acquired by Freshdesk ). He dropped out of Phd in ML to work with startups. He graduated from II IT H with specialization in theoretical comp science.\n\nHe has given tech talks at prestigious forums like PyData DC\, Fifth Elephant\, ICDCN\, POD C\, IIT Delhi\, IIIT Hyderabad and special interest groups like DLBLR. Mor e about him - https://www.linkedin.com/in/anuj-gupta-15585792/\n\n[Work fr om my past life](http://dblp.uni-trier.de/pers/hd/g/Gupta:Anuj)\n LAST-MODIFIED:20230810T072606Z LOCATION:Auditorium - MLR Whitefield\nIN ORGANIZER;CN="Anthill Inside":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/anthillinside/2017/schedule/synthetic-gradients-de coupling-layers-of-a-neural-nets-75WsziF58WKdqrJ3VKkSuN BEGIN:VALARM ACTION:display DESCRIPTION:Synthetic gradients: decoupling layers of neural nets. in Audi torium in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Sponsored session: AI on IA – a tutorial. DTSTART:20170729T093000Z DTEND:20170729T104500Z DTSTAMP:20260409T160148Z UID:session/KGTe9VpXp9P7qLEjVC2S1M@hasgeek.com SEQUENCE:2 CATEGORIES:Workshop,Intermediate CREATED:20170718T110036Z DESCRIPTION:The presentation will provide an overview of hardware and soft ware products for Deep learning that . The talk will also share details ab out Intel’s software optimization efforts to improve the performance of deep learning frameworks on Intel® architecture.\n\n### Speaker bio\n\nMu kesh Gangadhar works in Intel’s Software and Services Group\, where his role as a senior performance architect\, is focused on improving software applications in the areas of performance on Intel based platforms. His oth er interests include algorithm optimization for performance and power\, cl oud computing and artificial intelligence.\n LAST-MODIFIED:20230810T072606Z LOCATION:Breakout Room 1 - MLR Whitefield\nIN ORGANIZER;CN="Anthill Inside":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/anthillinside/2017/schedule/ai-on-ia-KGTe9VpXp9P7q LEjVC2S1M BEGIN:VALARM ACTION:display DESCRIPTION:Sponsored session: AI on IA – a tutorial. in Breakout Room 1 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:How deep is Deep Learning? DTSTART:20170729T100000Z DTEND:20170729T102000Z DTSTAMP:20260409T160148Z UID:session/Cgp7Z9b87xbh72sw3N5DEs@hasgeek.com SEQUENCE:2 CATEGORIES:Crisp talk,Intermediate CREATED:20170722T113717Z DESCRIPTION:+ Define and Explain Knowledge Tracing\n+ Explain the domain\, skills (knowledge components)\, funtoot platform and the dataset\n+ Discu ss and Explain Knowledge Tracing Models\n + Deep Knowledge Tracing (mod el and architecture)\n + DKT Applications: Discovering relationships an d interdepencies (pre-requisites)\n + Bayesian Knowledge Tracing\n + Performance Factor Analysis\n+ Analysis\, comparison and study of these m odels\n+ Discovering Limitations and Enhancements of BKT\n+ Discussion on the depth of deep learning and the knowledge tracing domain\n+ Conclusion\ n\n### Speaker bio\n\nAmar Lalwani\, Data Scientist @ funtoot\, is respons ible for research and development of funtoot's Brain. funtoot is a persona lised digital tutor in K-12 space for Math and Science. funtoot is activel y used by more than 1 lakh students and more than 130 schools across India .\n\nAmar Lalwani is also pursuing Ph.D. from IIIT-Bangalore in the area o f Machine Learning and Artificial Intelligence.\n LAST-MODIFIED:20230810T072606Z LOCATION:Banquet Hall - MLR Whitefield\nIN ORGANIZER;CN="Anthill Inside":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/anthillinside/2017/schedule/deep-learning-in-learn ing-education-Cgp7Z9b87xbh72sw3N5DEs BEGIN:VALARM ACTION:display DESCRIPTION:How deep is Deep Learning? in Banquet Hall in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Saving the Princess with Deep Learning. DTSTART:20170729T101500Z DTEND:20170729T103500Z DTSTAMP:20260409T160148Z UID:session/EGfzMVg4cnT598bdmDQKzc@hasgeek.com SEQUENCE:2 CATEGORIES:Crisp talk,Intermediate CREATED:20170719T043745Z DESCRIPTION:- Q Learning and DQN\, and why this is amazing!\n - Framework s we use\n - Problem definition\n - Breaking down the problem for our Ne ural Network\n - Designing the network\n - Training\, Validation and Tes ting\n - Demo\n - Hurdles\, Solutions and Optimisations\n\n### Speaker b io\n\nNavin is a graduate of the International Institute of Information Te chnology - Bangalore (IIIT-B) where he specialised in Machine Learning and AI. He has been a Pythonista for over half a decade and has conducted int roductory classes in Python for college students many times. He writes abo ut his experiments (in code and otherwise) at LifeOfNavin (http://lifeofna v.in) and Gradient Ascent (http://navinpai.github.io/ga) and actively cont ributes back to the open source community as well. He is currently working at Bloomreach\, a big data company focussed on the e-commerce sector. In his spare time\, he enjoys writing about himself from the third person per spective.\n LAST-MODIFIED:20230810T072606Z LOCATION:Auditorium - MLR Whitefield\nIN ORGANIZER;CN="Anthill Inside":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/anthillinside/2017/schedule/saving-the-princess-wi th-deep-learning-EGfzMVg4cnT598bdmDQKzc BEGIN:VALARM ACTION:display DESCRIPTION:Saving the Princess with Deep Learning. in Auditorium in 5 min utes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Evening beverage break DTSTART:20170729T102000Z DTEND:20170729T110500Z DTSTAMP:20260409T160148Z UID:session/4uR2dxHC8XMzzw99hZCDfC@hasgeek.com SEQUENCE:0 CREATED:20170616T072609Z DESCRIPTION:\n LAST-MODIFIED:20170724T122905Z LOCATION:Banquet Hall - MLR Whitefield\nIN ORGANIZER;CN="Anthill Inside":MAILTO:no-reply@hasgeek.com BEGIN:VALARM ACTION:display DESCRIPTION:Evening beverage break in Banquet Hall in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Evening beverage break DTSTART:20170729T103500Z DTEND:20170729T110500Z DTSTAMP:20260409T160148Z UID:session/gWyLoLhbMgLS9VVR89V8s@hasgeek.com SEQUENCE:0 CREATED:20170722T104011Z DESCRIPTION:\n LAST-MODIFIED:20170724T122757Z LOCATION:Auditorium - MLR Whitefield\nIN ORGANIZER;CN="Anthill Inside":MAILTO:no-reply@hasgeek.com BEGIN:VALARM ACTION:display DESCRIPTION:Evening beverage break in Auditorium in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Off The Record (OTR) session: "DL and image." DTSTART:20170729T105000Z DTEND:20170729T113500Z DTSTAMP:20260409T160148Z UID:session/FovyuHtVqv1pCK9KKTpSgk@hasgeek.com SEQUENCE:2 CATEGORIES:Full talk,Beginner CREATED:20170722T132256Z DESCRIPTION:The OTR will have participants from the field leading these di scussions.\n\n### Speaker bio\n\n\n LAST-MODIFIED:20230810T072606Z LOCATION:Breakout Room 1 - MLR Whitefield\nIN ORGANIZER;CN="Anthill Inside":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/anthillinside/2017/schedule/otr-dl-and-image-Fovyu HtVqv1pCK9KKTpSgk BEGIN:VALARM ACTION:display DESCRIPTION:Off The Record (OTR) session: "DL and image." in Breakout Room 1 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:AI in self driving vehicles: a practitioner's perspective. DTSTART:20170729T110500Z DTEND:20170729T115000Z DTSTAMP:20260409T160148Z UID:session/Ns7Z8id7ncfwSrz3RpyX4x@hasgeek.com SEQUENCE:2 CATEGORIES:Full talk,Beginner CREATED:20170722T115100Z DESCRIPTION:The talk will cover\n\n- Vision - CNNs and more\n- Driving pol icy - A good candidate for reinforcement learning\n- Data collection and q uality challenges\n- Simulators - transfer learning\n- Engineering challen ges - performance\, storage\n\n### Speaker bio\n\nSaad Nasser is a co-foun der at Ati Motors. He is an engineering prodigy who dropped out from schoo l at age 10 to pursue his own education. He has recently turned 15 and han dles technology at Ati where he works across autonomy\, power electronics and vehicle design.\n LAST-MODIFIED:20230810T072606Z LOCATION:Auditorium - MLR Whitefield\nIN ORGANIZER;CN="Anthill Inside":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/anthillinside/2017/schedule/ai-in-self-driving-veh icles-a-practitioners-perspective-Ns7Z8id7ncfwSrz3RpyX4x BEGIN:VALARM ACTION:display DESCRIPTION:AI in self driving vehicles: a practitioner's perspective. in Auditorium in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Apache MXNet: a highly memory efficient deep learning framework. DTSTART:20170729T110500Z DTEND:20170729T112500Z DTSTAMP:20260409T160148Z UID:session/A5zGLJt94BaeLZfgYZotyu@hasgeek.com SEQUENCE:2 CATEGORIES:Crisp talk,Intermediate CREATED:20170722T115419Z DESCRIPTION:What is MXNet\, where it came from\, who uses it\nHow it suppo rt both declarative and symbolic paradigms\nHow it minimize memory footpri nt trading of compute for memory\, impact of this tradeoff\nOther cool fea tures of MXNet like ease of programming\, ease of distributed programming\ , being able to access caffe\, torch layers\n\n### Speaker bio\n\nGirish P atil works as an Senior Solutions Architect for AWS. He focuses on deep le arning & traditional machine learning and big data projects. He helps many of India's highly successful start-ups to use these technologies. Girish is also a subject matter expert within Amazon on these technologies and re gularly participates global knowledge exchange programs. \n\nGirish's othe r interest include building internet scale applications. Girish also leads Developer Envagelism initiatives for AWS\, as well Innovation Pavilion in itiatives to promote young High Tech start-ups from India.\n LAST-MODIFIED:20230810T072606Z LOCATION:Banquet Hall - MLR Whitefield\nIN ORGANIZER;CN="Anthill Inside":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/anthillinside/2017/schedule/apache-mxnet-a-highly- memory-efficient-deep-learning-framework-A5zGLJt94BaeLZfgYZotyu BEGIN:VALARM ACTION:display DESCRIPTION:Apache MXNet: a highly memory efficient deep learning framewor k. in Banquet Hall in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:PyTorch demystified: why did I switch? DTSTART:20170729T112500Z DTEND:20170729T121000Z DTSTAMP:20260409T160148Z UID:session/Ff8rFCaCw7dZA2UV7Mb46E@hasgeek.com SEQUENCE:2 CATEGORIES:Full talk,Beginner CREATED:20170719T043810Z DESCRIPTION:- Computational Graphs [5 minutes]\n- Static and Dynamic graph s [5 minutes]\n- Introduction to PyTorch [10 minutes]\n - Variables\n - Functional\n - nn Module\n - Optim\n - AutoGrad\n- Examples [ 10 minutes]\n - Simple Neural Networks\n - Encoder Decoder Network\n - Stack-augmented Parser-Interpreter Neural Network\n- PyTorch and Caf fe2 [2 minutes]\n- How to visualize [5 minutes]\n- Benchmarking [1 minute] \n- QnA\n\n### Speaker bio\n\nI am working as an AI developer in [CoWrks]( www.cowrks.com). Here we work on a broad spectrum of use-cases in NLP and CV. I am currently focusing on NLP\, [particularly word representation usi ng image\, speech and predefined lexical structure](http://middleoutvector s.github.io/). In internet\, I go by hhsecond.\n LAST-MODIFIED:20230810T072606Z LOCATION:Banquet Hall - MLR Whitefield\nIN ORGANIZER;CN="Anthill Inside":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/anthillinside/2017/schedule/pytorch-demystified-wh y-did-i-switch-Ff8rFCaCw7dZA2UV7Mb46E BEGIN:VALARM ACTION:display DESCRIPTION:PyTorch demystified: why did I switch? in Banquet Hall in 5 mi nutes TRIGGER:-PT5M END:VALARM END:VEVENT END:VCALENDAR