MambularBase<T>

Base Classes · AiDotNet.NeuralNetworks.Tabular

Base class for Mambular (State Space Models for Tabular Data).

Implements: IParameterSource<T>

Known derived types: MambularClassifier<T>, MambularRegression<T>

For Beginners

Mambular is an alternative to transformers that:

  • Scales linearly: O(n) instead of O(n²) with sequence length
  • Has memory: Can remember information across the feature sequence
  • Is selective: Learns what to remember and what to forget

This makes it efficient for tabular data with many features.

How It Works

Mambular applies the Mamba architecture to tabular data:

  1. Features are embedded and treated as a sequence
  2. A stack of pre-norm residual Mamba layers, x + Mamba(RMSNorm(x)), processes the feature sequence. Each

Mamba layer is the full selective state space block of Gu and Dao (2023): input projection, causal depthwise conv1d, SiLU, the selective S6 scan with input-dependent delta, B and C, SiLU output gating and the output projection (MambaBlock).

  1. Final representation is used for prediction

Properties

Property Summary
long ParameterCount { get; }

Methods

Method Summary
Vector<T> GetParameters() Reads every parameter in traversal order.
void ResetState() Resets internal state.
void SetParameters(Vector<T> parameters) Restores every parameter, in the order GetParameters emitted them.
void UpdateParameters(T learningRate) Updates all parameters.