DerScherenmann/knnframework

Neural Network library in cpp

Neural Network library written in cpp

Usage:

First off design your network structure:

For example: 784 input neurons, 30 hidden neurons and 10 output neurons, all using the sigmoid function

std::vector<std::pair<int,int>> layers = {{784,Neuron::NeuronMode::SIGMOID},{30,Neuron::NeuronMode::SIGMOID}, {10,Neuron::NeuronMode::SIGMOID}};

Then create a network:

Network network = Network(layers);

Provide the network with adequate training data and parameters and start training:

network.train(trainingData,0.05,0.4,100);

If training has successfully completed, read the output values, after inputting some data:

std::vector<float> predictions = network.predict(testData);

Get the highest prediction, with wich you can read from predictions:

int index = network.highestPred(predictions);

Tips and Tricks:

You may experience some bugs or other weird stuff. Please open an issue or open a pull request if needed.

Training Data:

Training data in the MNIST database from http://yann.lecun.com/exdb/mnist/

Known bugs:

• Retraining the network after training finishes doesn't work