Crate neuro_divergent_training 

Neuro-Divergent Training Infrastructure

Comprehensive training system for neural forecasting models with advanced optimization, loss functions, and training strategies specifically designed for time series forecasting.

Features

• Advanced Loss Functions: Specialized forecasting losses (MAPE, SMAPE, MASE, CRPS, etc.)
• Modern Optimizers: Adam, AdamW, SGD, RMSprop with forecasting optimizations
• Learning Rate Schedulers: Exponential, step, cosine, plateau schedulers
• Unified Training Loop: Batch processing, gradient clipping, mixed precision
• Validation Framework: Cross-validation and model evaluation for time series
• Training Callbacks: Early stopping, checkpointing, progress tracking
• ruv-FANN Integration: Seamless integration with ruv-FANN neural networks

Example Usage

use neuro_divergent_training::*;
use ruv_fann::Network;

// Create a trainer with Adam optimizer and MAPE loss
let mut trainer = TrainerBuilder::new()
    .optimizer(AdamOptimizer::new(0.001, 0.9, 0.999))
    .loss_function(MAPELoss::new())
    .scheduler(ExponentialScheduler::new(0.001, 0.95))
    .build();

// Train the model
let result = trainer.train(&mut network, &training_data, &config)?;

Re-exports

pub use loss::*;

pub use optimizer::*;

pub use scheduler::*;

pub use metrics::*;

Modules

loss

Loss Functions for Neural Forecasting

metrics

Evaluation Metrics for Neural Forecasting

optimizer

Optimizers for Neural Forecasting

scheduler

Learning Rate Schedulers for Neural Forecasting

utils

Utility functions

Structs

CheckpointConfig

Checkpoint configuration

EpochMetrics

Metrics for a single epoch

LossAdapter

Adapter for integrating different loss functions with ruv-FANN

Network

A feedforward neural network

TimeSeriesMetadata

Metadata for individual time series

TrainingBridge

Bridge for integrating with ruv-FANN training algorithms

TrainingConfig

Training configuration

TrainingData

Core training data structure for time series

TrainingResults

Training results and metrics

Enums

CheckpointMode

DeviceConfig

Device configuration for training

NetworkError

Errors that can occur during network operations

TrainingError

Error types for training operations

Type Aliases

TrainingResult