LSTMNeuralNetwork<T>

Models & Types · AiDotNet.NeuralNetworks

Represents a Long Short-Term Memory (LSTM) Neural Network, which is specialized for processing sequential data like text, time series, or audio.

Inheritance: NeuralNetworkBase<T> → DeclaredModelLayoutBase<T> → SequenceModelLayoutBase<T> → LSTMNeuralNetwork<T>

For Beginners

An LSTM Neural Network is a special type of neural network designed for understanding sequences and patterns that unfold over time.

Think of an LSTM like a smart notepad that can:

  • Remember important information for long periods
  • Forget irrelevant details
  • Update its notes with new information
  • Decide what parts of its memory to use for making predictions

For example, when processing a sentence like “The clouds are in the sky”, an LSTM can:

  • Remember “The clouds” as the subject even after seeing several more words
  • Understand that “are” should agree with the plural “clouds”
  • Predict that “sky” might come after “in the” because clouds are typically in the sky

LSTMs are particularly good at:

  • Text generation and language modeling
  • Speech recognition
  • Time series prediction (like stock prices or weather)
  • Translation between languages
  • Any task where the order of data matters and patterns may span across long sequences

How It Works

Long Short-Term Memory networks are a special kind of recurrent neural network designed to overcome the vanishing gradient problem that traditional RNNs face. LSTMs have a complex internal structure with specialized “gates” that regulate the flow of information, allowing them to remember patterns over long sequences and selectively forget irrelevant information.

Example

var input = Tensor<float>.CreateRandom(new[] { 1, 20, 10 });
var trainX = Tensor<float>.CreateRandom(4, 8);
var trainY = Tensor<float>.CreateRandom(4, 2);
var result = new AiModelBuilder<float, Tensor<float>, Tensor<float>>()
    .ConfigureModel(new LSTMNeuralNetwork<float>())
    .Build(trainX, trainY);
var output = result.Predict(input);

Constructors

Constructor Summary
LSTMNeuralNetwork() Creates a new LSTM Neural Network with customizable loss and activation functions, using scalar activation functions.
LSTMNeuralNetwork(NeuralNetworkArchitecture<T> architecture, ILossFunction<T> lossFunction, IActivationFunction<T> outputActivation, IActivationFunction<T> forgetGateActivation, IActivationFunction<T> inputGateActivation, IActivationFunction<T> cellGateActivation, IActivationFunction<T> outputGateActivation, LSTMOptions options) Creates a new LSTM Neural Network with customizable loss and activation functions, using vector activation functions.
LSTMNeuralNetwork(NeuralNetworkArchitecture<T> architecture, ILossFunction<T> lossFunction, IVectorActivationFunction<T> outputVectorActivation, IVectorActivationFunction<T> forgetGateVectorActivation, IVectorActivationFunction<T> inputGateVectorActivation, IVectorActivationFunction<T> cellGateVectorActivation, IVectorActivationFunction<T> outputGateVectorActivation, LSTMOptions options) Creates a new LSTM Neural Network with customizable loss and activation functions, using vector activation functions.

Methods

Method Summary
ModelMetadata<T> GetModelMetadata() Gets metadata about the LSTM model.