DeepNeuralNetworkKernel<T>

Models & Types · AiDotNet.Kernels

Implements a deep (multi-layer) Neural Network kernel.

Implements: IKernelFunction<T>

For Beginners

This kernel corresponds to a deep neural network with multiple layers.

In each layer:

  1. Input from previous layer goes through the arc-cosine kernel transformation
  2. Output becomes input to the next layer

More layers = more compositional expressiveness, similar to deep neural networks. However, very deep kernels can have vanishing/exploding gradients just like deep NNs.

Constructors

Constructor Summary
DeepNeuralNetworkKernel(int order, double weightVariance, double biasVariance, int numLayers) Initializes a deep Neural Network kernel.

Properties

Property Summary
int NumLayers { get; } Gets the number of layers.

Methods

Method Summary
T Calculate(Vector<T> x1, Vector<T> x2) Calculates the deep Neural Network kernel value between two vectors.