EvoNet¶

EvoLib integration layer for EvoNet.

Provides an interface to EvoNet networks so it can be used inside EvoLib’s evolutionary pipeline. Supports configuration, mutation, crossover, and conversion to/from vector form.

class evolib.representation.evonet.EvoNet[source]¶

Bases: ParaBase

Wrapper class for EvoNet.

Responsibilities: - Build and configure neural networks from YAML/typed configs - Provide mutation (weights, biases, activations, structure) - Provide crossover at weight/bias level (no structural crossover) - Expose network parameters as flat vectors for integration

apply_config(cfg)[source]¶

Initializes parameters from a configuration object.

Return type:: None

calc(input_values)[source]¶

Return type:: list[float]

crossover_with(partner)[source]¶

Perform crossover on weights and biases if topologies are compatible.

Structural crossover is not supported.

Return type:: None

get_biases()[source]¶

Return network biases (non-input neurons).

Return type:: ndarray

get_status()[source]¶

Return type:: str

get_vector()[source]¶

Return a flat vector containing all weights and biases.

Return type:: ndarray

get_weights()[source]¶

Return network weights in the canonical order defined by Nnet.

Return type:: ndarray

mutate()[source]¶

Applies mutation to the parameters.

Return type:: None

plot(name, engine='neato', labels_on=True, colors_on=True, thickness_on=False, fillcolors_on=False)[source]¶

Prints the graph structure of the EvoNet.

Parameters:

name (str) – Output filename (without extension).
engine (str) – Layout engine for Graphviz.
labels_on (bool) – Show edge weights as labels.
colors_on (bool) – Use color coding for edge weights.
thickness_on (bool) – Adjust edge thickness by weight.
fillcolors_on (bool) – Fill nodes with colors by type.

Return type:

None

print_status()[source]¶

Return type:: None

set_biases(biases)[source]¶

Set network biases; length must match num_biases.

Return type:: None

set_vector(vector)[source]¶

Split a flat vector into weights and biases and apply them to the network.

Return type:: None

set_weights(weights)[source]¶

Set network weights; length must match num_weights.

Return type:: None

update_mutation_parameters(generation, max_generations, diversity_ema=None)[source]¶

Optional: Override in subclasses that support strategy-dependent mutation control.

Return type:: None