Guidance to start with Neural Networks

I don’t want someone interested in neural networks to struggle like me, so I would like to share about my learning path, hope it will make your learning simple.

Image made by Fjodr van veen

First of all, try to finish this online book, which will give strong basics about neural networks. yes it is big, but remember more patience makes your basics more stronger.

Now after getting confidence of what neural network is and how it works, Watch these videos to recall everything you read till now.

For now, you will have proper understanding about neural network.
To go forward, read:

this will give a lot of information about convolution neural networks which are giving lot of surprising outputs in image processing.

you can also look into below link, It visualizes what your convolution neural network is learning (see this only only after completing above tasks), try to experiment by modifying the models

Now, read about recurrent neural networks (RNN) that are backbone of modern text processing

After this, read about the beauty of a new trending block of recurrent neural network ie., LSTM (Long Short Term Memory)

To go deep into RNN’s read: (It also has a great visualization to understand)

Try to run below code, it is very interesting output from RNN’s written by Karpathy

After finishing everything mentioned above, watch this talk (By Naftali Tishby) as many times as you can, Seriously it is has lot of information which is very difficult to understand at once.

Now lets get into something more interesting, GAN(General Adversarial network). It is super awesome concept, here one network evaluates another network.

Start with Vanilla GAN, try to understand the code and watch the output:

Know about Variational Auto Encoders(VAE), Same like in GAN start with Vanilla VAE:

You can also read about few more concepts in deep learning:

  • Restricted Boltzmann Machines
  • Deep belief networks

Few interesting things in web:

  • 3D face Reconstruction from single Image:
  • Draw together with RNN:
  • Teach Machine using your camera live in the browser:
  • Rubix Cube Solver:
  • Behind scenes of Allo (Smart Chat app made by Google):
  • Behind scenes of Auto reply in Gmail (have few wow factors):
  • Lot of good examples from PyTorch Tutorials:
  • Colorizing Black & White Photos using Deep Learning:

Few Interesting Research papers:


Popular frameworks:

(Note all the information above is shortlisted by myself after reading many things like this from web, comment below if you know any other good reads about neural networks)