Here we cover the following topics
- Saving and restoring trained models
- L1 and L2 regularization
- Using different flavors of RNNs like LSTM, GRU
- Using RNN for different usage like regression and MNIST handwritten digit classification
- The programs are self contained for ease of understanding
- Using dropout in images classification
- Setting up bidirectional and multilayer RNNs.
- Testing out different activation functions because numerical regression is different from other tasks like classification and thus demands a bit different activation function
- L2 regularization
- filename : linear_regression.py
- L2 regularization
- Using dropout in image classification
- Saving and restoring models
- Using MNIST images from torchvision
- Moving models to specific device (GPU / CPU)
- Setting up bidirectional and multilayer RNNs.
- filename : mnist_classifier.py
- Same as above but uses Gated Recurrent Unit (GRU)
- filename : mnist_classifier.py
- Learning to use different activation functions
- filename : sine_approximation.py