RedFT/nnet

A small neural network library

Synopsis

This is a simple N-layer Neural Network library. [This project uses Python 3]

Motivation

I am following Stanford's CS231n course (Convolutional Neural Networks for Visual Recognition). I decided I would build a small Neural Network library as I progress through this course.

Code Example

Below is an example of a two layer neural network. You can find similar code in test/run_tests.py.

nn = nnet.NeuralNetwork(input_size)

nn.add_layer(nnet.FullyConnectedLayer(
    initialization_type="xavier_relu", 
    output_size=100, 
    pass_type="test|train"))
nn.add_layer(nnet.ActivationLayer(
    "relu", 
    pass_type="test|train"))
nn.add_layer(nnet.FullyConnectedLayer(
    initialization_type="xavier_relu", 
    output_size=10, 
    pass_type="test|train"))
nn.add_layer(nnet.SoftmaxLayer(
    pass_type="test"))
nn.add_layer(nnet.LossLayer(
    "softmax", 
    pass_type="train"))

nn.train(training_data, training_labels, 
    testing_data, testing_labels)

Documentation

Documentation is minimal.

To generate html documentation, you must first install sphinx.

$ pip3 install sphinx

Change into the docs directory. You can then generate html documentation by running:

$ make html

To view the html documentation in your web browser (i.e. firefox), you can run:

$ firefox build/html/index.html

Install Dependencies

From the project's root directory run:

$ pip3 install -r requirements.txt

Test

Change into the test directory. To run the test script, you need to make sure you have the Cifar10 dataset in the data directory. Just run:

$ chmod +x get_cifar10_dataset.sh
$ ./get_cifar10_dataset.sh

Then from the project's root directory:

$ nosetests -v

License

This code in this project is released under the GPL License.