SimpleNeuralNetwork<T>

Models & Types · AiDotNet.Prototypes

Simple 2-layer neural network for prototype validation. Demonstrates GPU acceleration through vectorized operations.

How It Works

This is a minimal neural network for Phase A prototype validation. It includes:

  • Input layer → Hidden layer (with ReLU activation)
  • Hidden layer → Output layer (linear)
  • Backpropagation
  • Adam optimizer integration

Constructors

Constructor Summary
SimpleNeuralNetwork(int inputSize, int hiddenSize, int outputSize, int? seed) Initializes a new instance of the SimpleNeuralNetwork.

Properties

Property Summary
long ParameterCount { get; } Gets the total number of parameters in the network.

Methods

Method Summary
PrototypeVector<T> ComputeLossGradient(PrototypeVector<T> predicted, PrototypeVector<T> target) Computes gradient of MSE loss.
PrototypeVector<T> Forward(PrototypeVector<T> input) Forward pass through the network.
PrototypeVector<T> GetParameters() Gets all parameters as a single flattened vector.
T ComputeLoss(PrototypeVector<T> predicted, PrototypeVector<T> target) Computes mean squared error loss.
ValueTuple<PrototypeVector<T>, PrototypeVector<T>, PrototypeVector<T>, PrototypeVector<T>> Backward(PrototypeVector<T> outputGradient) Backward pass (computes gradients).
void SetParameters(PrototypeVector<T> parameters) Sets all parameters from a single flattened vector.