System requrements:
- Python 3.6,
- python-mnist package
- numpy package
To simplify the obtaining dataset, please use get_training_data script.
pip3 install python-mnist numpy
python3 get_training_data.py
neural_network.py implemented NeuralNetwork class and related functionality. Run example:
python3 neural_network.py train-images/ 10 0.05 0.01 100
Inputs:
- folder with training data (in .gz format)
- stop criterion by the number of eras - stop criterion for minimizing cross-entropy - learning Speed - the number of neurons in the hidden layer