Deep Learning notes.
- Setup - Software, Hardware: Software and Hardware setup for DL.
- Introduction to Python: A Crash course on python.
- Introduction to Numpy: A Crash course on Numpy.
- Understanding bias-Variance tradeoff
- Introduction to Neural networks: How does NN classify non linear data?. Is it a stacked ensemble of linear classifiers?.
- Understanding gradient flows: Understanding Backpropagation using gradient flow.
- Understanding convolution: Implementing the convolution from scratch, and comparing it with tensorflow implementation.
- Understanding convolution part 2: Compare convolutions using a predefined filter like gabor with a learnt filter using backpropagation.
- Transfer Learning: Learn to use inception Resnet V2 architecture for transfer learning.
- CNN Architectures:
- Understanding Resnet: Paper summary of Deep Residual Learning for Image Recognition.
- Wide resnet on CIFAR 10: Classification of CIFAR 10 dataset using wide resnet. Contains input data augmentaion using the Tensorflow Dataset API, slim based wide resnet implementaion, achieving an accuracy of ~93.23%.
- U-net Segmentation of Medical images: A TF-slim implementaion of U-net arachitecture for segmenting White blood cells on Medical images.
- Recurrent Neural Network Character level RNN implementation and experiments on state variables.
- Exploding gradients in RNN: Understanding the exploding and vanishing gradient problem in RNNs.
The above code was developed on Linux using Python 3.5, Tensorflow 1.8.