RBMLayer<T>

Layers · AiDotNet.NeuralNetworks.Layers

Represents a Restricted Boltzmann Machine (RBM) layer for neural networks.

Inheritance: LayerBase<T> → RBMLayer<T>

Implements: IShapeContract

For Beginners

An RBM layer is like a feature detector that can learn patterns in data.

Imagine you have a set of movie ratings:

  • The visible layer represents the actual ratings
  • The hidden layer represents abstract features (e.g., “likes action”, “prefers comedy”)
  • The RBM learns to connect ratings to these abstract features

RBM layers are useful for:

  • Finding underlying patterns in data
  • Reducing the dimensionality of data
  • Initializing weights for deep neural networks

How It Works

An RBM layer is a stochastic neural network layer that learns a probability distribution over its inputs. It consists of a visible layer and a hidden layer with no connections between nodes within the same layer.

Constructors

Constructor Summary
RBMLayer(int hiddenUnits, IActivationFunction<T> scalarActivation) Lazy ctor — visible-unit count is resolved from input.Shape[^1] on the first Forward call.
RBMLayer(int visibleUnits, int hiddenUnits, IActivationFunction<T> scalarActivation) Initializes a new instance of the RBMLayer class with vector activation.
RBMLayer(int visibleUnits, int hiddenUnits, IVectorActivationFunction<T> vectorActivation) Initializes a new instance of the RBMLayer class with vector activation.

Properties

Property Summary
bool IsInitialized { get; }
bool SupportsTraining { get; } Indicates whether this layer supports training.

Methods

Method Summary
Dictionary<string, string> GetParameterRoles() Returns parameter roles for per-role optimizer configuration (e.g., weight decay exemption for biases).
IReadOnlyList<OutputAxisContract> OutputAxesFor(int inputRank)
IReadOnlyList<Tensor<T>> GetTrainableParameters() Returns all trainable parameter tensors marked with [TrainableParameter].
void ClearGradients()
void ResetState()
void SetTrainableParameters(IReadOnlyList<Tensor<T>> parameters) Replaces trainable parameter tensors (e.g., with ParameterBuffer views).
void TrainWithContrastiveDivergence(Vector<T> input, T learningRate, int kSteps) Trains the RBM with a tensor whose last axis is the visible-unit axis.
void UpdateParameters(T learningRate) Updates the layer’s parameters using either standard backpropagation or contrastive divergence.
void ZeroGrad() Clears all gradient fields discovered by convention ({paramName}Gradient).