My own well-commented implementation of Feed-Forward Neural Network and Back Propagation algorithm on pure python with numpy.
Features:
- Adoptive learning rate
- Matrixes operation to improve efficiency.
- Flexible activation-function (Abstract class to implement any function. Sigmoid and ReLU are built-in)
- Flexible loss-function (Abstract class to implement any function. MSE and LogLoss are built-in)
- Built-in cross-validation.
File NeuralInUse.ipynb contains an example of using the Neural Network. Maybe interesting for the ones who are styding NN.