Convolution Neural Networks without any frameworks

prakhar srivastava (~prakhar91)


31

Votes

Description:

Convolution Networks - Framework = Vision in vanilla python.

This workshop aims at providing the attendees an experience of implementing convolution neural from scratch without any big framework working in the backend supplementing the need for computation. This would give the attendees an overall understanding of what are Convolution Neural Networks and why do they work so exceedingly well!

image

One does not simply code in vanilla python.

What can you expect from this workshop!

  1. You'll understand what are convolution neural networks
  2. Why they work so well on image data?
  3. All the different implementation of Convolution network and how they improve the vanilla network
  4. What are the best ways to implement convolution network on a given data

What this workshop is not!

  1. Just another workshop telling you to use frameworks
  2. Maths will not be looked over. (It's important)
  3. This workshop is not any other university lecture where you'll not understand anything.

I find this image to be so apt given all the abstraction provided by frameworks

image

Prerequisites:

  1. Command over Python
  2. Familiarity with Numpy and basic math packages
  3. Intermediate Mathematics
  4. Familiarity with algorithms common in machine learning

Content URLs:

Coming up soon (related to this workshop)

Speaker Info:

Hello World! I'm Prakhar Srivastava, junior year undergrad, a deep learning enthusiast who loves mathematics and astronomy. I've been exploring machine learning/deep learning for about 2 years now and fiddling with the basic mathematics and scratch implementations always excite me. I'm currently mentor of deep learning in a Delhi based startup Greatech Soft Solutions and interning at Startup labs and a Google Summer of Code '18 student under the organization OpenAstronomy.

Speaker Links:

  1. http://prsr.me
  2. https://www.linkedin.com/in/prakharcode
  3. https://github.com/prakharcode

Section: Data science
Type: Workshops
Target Audience: Intermediate
Last Updated: