-
To check the gradient from numerical and backprop, simply run:
python check_gradient.py -
To train the neural network, simply run:
python neuralnet.py
Change config accordingly based on different quetions. The test accuracy will be printed, and the train/validation loss/accuracy will be saved in a .pkl file.
- To get the loss/accuracy plot, simply run:
python plot_<question number>.py
Each question has one file for plot purpose. Remember to change the filename to generated .pkl file.
- In case the test accuracy is not recorded, we still can get the test accuracy after training. Simply run:
python get_test_accuracy.py
Remember to change the filename to the .pkl file.