MachineLearning/ann library
🧠Artificial Neural Network (ANN)
A minimal, production-minded feed-forward neural network API. This class exposes a generic, well-documented interface for building multi-layer perceptrons with configurable layer sizes, activation functions, and training parameters. The implementation favors clarity, numerical safety, and explicitness suitable for engineering evaluation and extension.
Contract:
- Input: feature matrix X (n x m) as List<List
- Output: trained weights and a
predictmethod which returns model outputs. - Errors: throws ArgumentError for invalid shapes and StateError if used before fit.
Notes: This implementation is intentionally lightweight (no GPU) and designed to be a maintainable, auditable building block rather than a high-performance library.