The Fifth Elephant 2015

A conference on data, machine learning, and distributed and parallel computing

Introduction to Deep Learning

Submitted by Bargava Subramanian (@barsubra) on Monday, 18 May 2015

videocam_off

Technical level

Intermediate

Section

Workshop

Status

Confirmed & Scheduled

View proposal in schedule

Vote on this proposal

Login to vote

Total votes:  +20

Objective

In fields like computer vision, speech recognition and natural language processing, deep learning has produced state-of-art results. And they are showing lot of promise in other fields too.

This workshop will provide an introduction to deep learning. It would cover some of the common deep learning architectures, advantages and concerns, along with some hands-on.

Description

  • What is deep learning?
  • Motivation: Some use cases where it has produced state-of-art results
  • Supervised learning (multi-layer perceptron, deep convolution networks)
  • Unsupervised learning (autoencoders) time permitting
  • Overview of a few libraries and the impact of GPUs (Some practical thoughts on hardware and software)
  • Hands-on modeling a simple classification problem

We would be using Python-based libraries to do the hands-on.

The data and software requirements would be posted onto the github repository one week prior to the workshop.

The repository for this workshop:
https://github.com/rouseguy/intro2deeplearning/

EDIT(13July2015): Software and data installations are posted on the repository. Please install them prior to the workshop.

Requirements

If you’ve never built a predictive model, this is probably NOT the ideal workshop to attend. Having said this, the attendee should, at the bare minimum, understand the following terms:
1. Bias and Variance
2. Train, Test and Validation sets
3. Cross-validation, grid-search, hyperparameter optimization
4. Measuring model accuracy (Precision, Recall, F1 score, Area Under Curve)
5. Supervised and Unsupervised learning

Also, to follow the hands-on, the attendee should have had some programming experience (Reading files, performing some data manipuation on them, loops, conditional expressions)

We would be using Python stack for the hands-on. Knowing Python is a plus, but not mandatory.

FAQs

  1. Question we repeatedly get asked: Do I need a laptop with GPU? Do I need a powerful processing machine? Do I need a lot of RAM? Do I need a cloud compute account? Fret not. We would do something really small-scale. A laptop with 4 GB RAM should suffice.

  2. I use a Windows machine? We haven’t used a Windows machine in a while and so, would be a challenge to support bugs/issues in Windows. We recommend installing a linux VM.

  3. Data and the Software libraries? Please check the github repository a week before the workshop. Data and the library installation instructions would be posted there.

  4. I dont know Python! That’s perfectly fine. The concepts aren’t tied to any language. The constructs used in the code would be as simple as possible.

Speaker bio

  • Bargava Subramanian is a Senior Statistician(Data Scientist) at Cisco Systems, India. He has a Masters from University of Maryland, College Park, USA.

  • Raghotham is a full-stack developer at RedMart. He has a Masters from BITS, Pilani. In his previous role, he was instrumental in architecting an analytics platform for a wearable devices company.

Links

Slides

https://github.com/rouseguy/intro2deeplearning

Comments

Login with Twitter or Google to leave a comment