GaussianMixtureClassifierLayer<T>
Layers · AiDotNet.NeuralNetworks.Layers
A Gaussian mixture classifier: one diagonal-covariance Gaussian mixture per class, emitting each class’s
log-likelihood log p(x | class). Under equal class priors a softmax over the output is the Bayes posterior.
Inheritance: LayerBase<T> → GaussianMixtureClassifierLayer<T>
Implements: IShapeContract
For Beginners
Each class gets a few “clouds” of points describing where its examples usually lie. A new example is scored by how well it fits each class’s clouds, and the best-fitting class wins.
How It Works
This is the statistical pattern recognition classifier of Tzanetakis and Cook 2002 (section IV): a mixture of
components Gaussians with diagonal covariance per class, initialized with K-means and fitted with EM
(Fit). One component is their “simple Gaussian” classifier. The parameters are also differentiable, so
the layer can be trained discriminatively.
Constructors
| Constructor | Summary |
|---|---|
GaussianMixtureClassifierLayer(int classes, int components, int features) |
Creates the layer with unit-variance components at the origin and equal weights. |
Properties
| Property | Summary |
|---|---|
bool SupportsTraining { get; } |
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 ResetState() |
|
void SetTrainableParameters(IReadOnlyList<Tensor<T>> parameters) |
Replaces trainable parameter tensors (e.g., with ParameterBuffer views). |
void ZeroGrad() |
Clears all gradient fields discovered by convention ({paramName}Gradient). |