neural-network-framework

Build a Micro Neural Network Framework (like Pytorch or Tensorflow) using only Numpy

Layers:

Main Blocks:

- Fully Connected
- Convolutional
- RNN

CNN Blocks

- Pooling
- Flatten

RNN Blocks:

- Dropout
- Batch normalization

Activation Functions:

- ReLU
- Soft max
- Sigmoid

Optimization:

Optimizers:

- Stochastic Gradient Descent
- SGD With Momentum
- Adam

Loss Functions:

- Cross Entropy Loss

Regularizers:

- L1
- L2