MNIST image classifier and easy-to-use Java ANN library, version 1.2. Create, save, load, and train neural networks.
- Creating networks given layer size parameters
Network net = new Network(784,100,10);
- Loading networks from text files
Network.loadNetworkFromResource("foo.txt");
- Saving networks at text files
net.saveNetworkAtResource("foo.txt");
- Training networks from MNIST dataset
net.trainFromMnistData("data/train-images.idx3-ubyte", "data/train-labels.idx1-ubyte", eta, start, end, epochs, bw);
- Testing networks from MNIST dataset
net.testAll("data/t10k-images.idx3-ubyte", "data/t10k-labels.idx1-ubyte", start, end, showMistakes, bw);
- Network at network.txt that can classify MNIST images
- Image prompt, applet, and classifier to showcase library
- Implement other types of neural nets
- Add support for other datasets/applications
- GUI to facilitate usage of library
- Optimize and implement more efficient algorithms
Written by Reedit Shahriar in January 2018. Licensed under MIT License.
- Used code from this repo to open MNIST images and read data.