ConvNetJSJava

ConvNetJS is a Javascript implementation of Neural networks, together with nice browser-based demos. Do check it out ConvNetJS

While, this fork is, well, a java implementation I transliterated while learning Neural Networks 😁.

It's supposed to be supporting:

• Common Neural Network modules (fully connected layers, non-linearities)
• Classification (SVM/Softmax) and Regression (L2) cost functions
• Ability to specify and train Convolutional Networks that process images
• An experimental Reinforcement Learning module, based on Deep Q Learning

For much more information, um, I have no determination on maintaining this project (and it seems, the original fork is also not maintained) feel free to contact me at Discord. Or consider the main page at convnetjs.com

Note: I am not actively maintaining ConvNetJS anymore because I simply don't have time. I think the npm repo might not work at this point.

Online Demos

I wish I could write some swing demos...

• Convolutional Neural Network on MNIST digits
• Convolutional Neural Network on CIFAR-10
• Toy 2D data
• Toy 1D regression
• Training an Autoencoder on MNIST digits
• Deep Q Learning Reinforcement Learning demo
• Image Regression ("Painting")
• Comparison of SGD/Adagrad/Adadelta on MNIST

Example Code

Here's a minimum example of defining a 2-layer neural network and training it on a single data point:

    // species a 2-layer neural network with one hidden layer of 20 neurons
    var layer_defs = new VP.VPL();
    // input layer declares size of input. here, 2-D data
    // ConvNetJS works on 3-Dimensional volumes (sx, sy, depth), but if you're not dealing with images
    // then the first two dimensions (sx, sy) will always be kept at size 1
    layer_defs.push("type", "input", "out_sx", 1, "out_sy", 1, "out_depth", 2);
    // declare 20 neurons, followed by ReLU (rectified linear unit non-linearity)
    layer_defs.push("type", "fc", "num_neurons", 20, "activation", "relu");
    // declare the linear classifier on top of the previous hidden layer
    layer_defs.push("type", "softmax", "num_classes", 10);

    var net = new Net();
    net.makeLayers(layer_defs);

    // forward a random data point through the network
    var x = new Vol(0.3, -0.5);
    var prob = net.forward(x);

    // prob is a Vol. Vols have a field .w that stores the raw data, and .dw that stores gradients
    System.out.println("probability that x is class 0, " + prob.w.get(0)); // prints 0.50101

    var trainer = new Trainer(net, "learning_rate", 0.01, "l2_decay", 0.001);
    for (int i = 0; i < 100; i++) {
        trainer.train(x, 0); // train the network, specifying that x is class zero
    }

    var prob2 = net.forward(x);
    System.out.println("probability that x is class 0, " + prob2.w.get(0));
    // now prints 0.50374, slightly higher than previous 0.50101, the networks
    // weights have been adjusted by the Trainer to give a higher probability to
    // the class we trained the network with (zero)

and here is a small Convolutional Neural Network if you wish to predict on images:

    var layer_defs = new VP.VPL();
    layer_defs.push("type", "input", "out_sx", 32, "out_sy", 32, "out_depth", 3); // declare size of input
    // output Vol is of size 32x32x3 here
    layer_defs.push("type", "conv", "sx", 5, "filters", 16, "stride", 1, "pad", 2, "activation", "relu");
    // the layer will perform convolution with 16 kernels, each of size 5x5.
    // the input will be padded with 2 pixels on all sides to make the output Vol of the same size
    // output Vol will thus be 32x32x16 at this point
    layer_defs.push("type", "pool", "sx", 2, "stride", 2);
    // output Vol is of size 16x16x16 here
    layer_defs.push("type", "conv", "sx", 5, "filters", 20, "stride", 1, "pad", 2, "activation", "relu");
    // output Vol is of size 16x16x20 here
    layer_defs.push("type", "pool", "sx", 2, "stride", 2);
    // output Vol is of size 8x8x20 here
    layer_defs.push("type", "conv", "sx", 5, "filters", 20, "stride", 1, "pad", 2, "activation", "relu");
    // output Vol is of size 8x8x20 here
    layer_defs.push("type", "pool", "sx", 2, "stride", 2);
    // output Vol is of size 4x4x20 here
    layer_defs.push("type", "softmax", "num_classes", 10);
    // output Vol is of size 1x1x10 here

    var net = new Net();
    net.makeLayers(layer_defs);

    // helpful utility for converting images into Vols is included
    Image image = ImageIO.read(***);
    var x = VolUtil.img_to_vol(image, true);
    var output_probabilities_vol = net.forward(x);

Getting Started

Hey! Dont use this project consider using ConvnetJS instead! Neither is it maintained, nor anyone can help. A Getting Started tutorial is available on main page.

The full Documentation can also be found there.

Must have:

allprojects {
    repositories {
        maven { url 'https://jitpack.io' }
    }
}

To get the library only:

testImplementation 'com.github.Minecraftian14:convnetjs:0.1.0'

To get the library and a few utilities:

testImplementation 'com.github.Minecraftian14:convnetjs:0.1.0:featured'

These utilities contains a few extra classes to make things slightly easier to use. For instance, the same example from above can be written as:

LayerVPL layer_defs = new LayerVPL();
layer_defs.input(2);
layer_defs.fc(20).activation("relu");
layer_defs.softmax(10);

Net net = new Net();
net.makeLayers(layer_defs);

Trainer trainer = new Trainer(net, new TrainerVP().method("sgd").learning_rate(0.01).l2_decay(0.001));

License

MIT