Jeras

Jeras is a high-level, fluent API for neural networks written in Java and modeled after Keras. It is designed to support Keras objects, such as:

• Optimizers
• Activations
• Layers
• Sequential models
• and more!

Code example

  // Build model architecture
  Sequential model = new Sequential();
  model.add(new Input(2).withBiasInitializer(ONES));
  model.add(new Dense(4).withBiasInitializer(CONSTANT.withValue(16)));
  model.add(new Dense(1));

  // Set training parameters
  model.compile()
          .withOptimizer(new Gd()
                  .withNesterov()
                  .withMomentum(0.2)
                  .withLearningRate(0.1))
          .withLossFunction(MSE);

  double[] y = model.predict(new double[]{1,2});
  System.out.println(Arrays.toString(y));

Current Status

Work-in-progress! Check out components here