Crate neuro_divergent_models 

Neuro-Divergent Models

A comprehensive neural forecasting library built on top of ruv-FANN, providing state-of-the-art time series forecasting models for production use.

This library implements 27+ neural forecasting models inspired by NeuralForecast, optimized for Rust’s performance and safety guarantees.

Features

• Recurrent Models: RNN, LSTM, GRU with temporal state management
• Transformer Models: Multi-head attention, TFT, and advanced architectures
• Linear Models: DLinear, NLinear with decomposition
• Specialized Models: NBEATS, TimesNet, and domain-specific architectures
• Production Ready: Type-safe, memory-efficient, and scalable

Quick Start

use neuro_divergent_models::{NeuralForecast, models::LSTM, LSTMConfig};
use neuro_divergent_models::data::TimeSeriesDataFrame;

// Create LSTM model
let lstm_config = LSTMConfig::default_with_horizon(24)
    .with_architecture(128, 2, 0.1)
    .with_training(1000, 0.001);

let lstm = LSTM::new(lstm_config)?;

// Create forecasting pipeline
let mut nf = NeuralForecast::new()
    .with_model(Box::new(lstm))
    .build()?;

// Train and forecast
nf.fit(train_data)?;
let forecasts = nf.predict()?;

Re-exports

pub use errors::NeuroDivergentError;

pub use errors::NeuroDivergentResult;

pub use foundation::BaseModel;

pub use foundation::NetworkAdapter;

pub use foundation::ModelConfig;

pub use foundation::TimeSeriesInput;

pub use foundation::ForecastOutput;

pub use foundation::ValidationConfig;

pub use data::TimeSeriesDataFrame;

pub use data::ForecastDataFrame;

pub use data::TimeSeriesSchema;

pub use forecasting::NeuralForecast;

pub use config::LSTMConfig;

pub use config::RNNConfig;

pub use config::GRUConfig;

pub use config::TrainingConfig;

pub use config::PredictionConfig;

pub use config::CrossValidationConfig;

Modules

activations

Activation functions for neural networks

config

Configuration structures for neural forecasting models

data

Data structures for time series handling

errors

Error types for neuro-divergent models

forecasting

Main forecasting interface for neuro-divergent models

foundation

Core foundation traits and structures for neural forecasting models

layers

Neural network layers for forecasting models

models

Neural forecasting model implementations

recurrent

Recurrent Neural Network Models

utils

Utility functions for neural forecasting models

Structs

Layer

Represents a layer of neurons in the neural network

Network

A feedforward neural network

Neuron

Represents a single neuron in the neural network

Enums

ActivationFunction

Activation functions available for neurons

Traits

Float

Generic trait for floating point numbers