Runners and Trainers#

This module provides various running and training algorithms for various neural networks.

The supported training algorithms include

  • offline training methods, like ridge regression, linear regression, etc.

  • online training methods, like recursive least squares (RLS, or Force Learning), least mean squares (LMS), etc.

  • back-propagation learning method

  • and others

The supported neural networks include

  • reservoir computing networks,

  • artificial recurrent neural networks,

  • and others.

Base RNN Runner#

RNNRunner(target[, jit])

Structural Runner for Recurrent Neural Networks.

Base RNN Trainer#

RNNTrainer(target, **kwargs)

Structural Trainer for Models with Recurrent Dynamics.

Online RNN Trainer#

OnlineTrainer(target[, fit_method])

Online trainer for models with recurrent dynamics.

ForceTrainer(target[, alpha])

Force learning.

Offline RNN Trainer#

OfflineTrainer(target[, fit_method])

Offline trainer for models with recurrent dynamics.

RidgeTrainer(target[, beta])

Trainer of ridge regression, also known as regression with Tikhonov regularization.

Back-propagation Trainer#

BPTT(target, loss[, optimizer, ...])

The trainer implementing back propagation through time (BPTT) algorithm for recurrent neural networks.

BPFF(target, **kwargs)

The trainer implementing back propagation algorithm for feedforward neural networks.