src.canns.trainer.sanger

Sanger’s rule for sequential principal component extraction.

Classes

SangerTrainer

Sanger's rule (Generalized Hebbian Algorithm) for multiple PC extraction.

Module Contents

class src.canns.trainer.sanger.SangerTrainer(model, learning_rate=0.01, normalize_weights=True, weight_attr='W', compiled=True, **kwargs)

Bases: src.canns.trainer._base.Trainer

Sanger’s rule (Generalized Hebbian Algorithm) for multiple PC extraction.

Extends Oja’s rule with Gram-Schmidt orthogonalization to extract multiple principal components. Each neuron learns to be orthogonal to all previous ones.

Learning Rule:

ΔW_i = η * (y_i * x - y_i * Σ_{j≤i} y_j * W_j)

where:

• W_i is the i-th neuron’s weight vector

• y = W @ x is the output vector

• The sum enforces orthogonality (Gram-Schmidt process)

This allows sequential extraction of orthogonal principal components, with neuron i converging to the i-th principal component.

Reference:

Sanger, T. D. (1989). Optimal unsupervised learning in a single-layer linear feedforward neural network. Neural Networks, 2(6), 459-473.

Initialize Sanger trainer.

Parameters:

• model (src.canns.models.brain_inspired.BrainInspiredModel) – The model to train (typically LinearLayer)

• learning_rate (float) – Learning rate η for weight updates

• normalize_weights (bool) – Whether to normalize weights to unit norm after each update

• weight_attr (str) – Name of model attribute holding the connection weights

• compiled (bool) – Whether to use JIT-compiled training loop (default: True)

• **kwargs – Additional arguments passed to parent Trainer

predict(pattern, *args, **kwargs)

Predict output for a single input pattern.

Parameters:

pattern – Input pattern of shape (input_size,)

Returns:

Output pattern of shape (output_size,)

train(train_data)

Train the model using Sanger’s rule.

Parameters:

train_data (collections.abc.Iterable) – Iterable of input patterns (each of shape (input_size,))

compiled = True

learning_rate = 0.01

normalize_weights = True

weight_attr = 'W'