Genome¶

Main Object:¶

class abrain.Genome(key=None)[source]¶

Genome class encoding a CPPN either for ES-HyperNEAT or for direct use.

A simple collection of Node/Link. Can only be created via random() init, crossover() or copy() (with the help of the nested Data structure)

class Data(seed: int, cdata: dict, labels: str | Collection[str] | None, genome_ids: bool = True, innovations: bool = True, key=None)[source]¶

Long-lived data structure encapsulating the various elements needed throughout evolution

cdata: dict¶: CPPN creation data (shape, functions, …)

rng: Random¶: Source of randomness

id_manager: Innovations | IDManager¶: Either an innovations database (allowing crossover but slightly more expensive) or a monotonic index-based allocator (for mutations only)

gid_manager: GIDManager | None¶: (Optional) Manager for genome identifiers and lineage

labels: List[str] | None¶: (Optional) Labels for pretty-printing the input/output nodes

static create_for_generic_cppn(inputs: int, outputs: int | Collection[str], labels: str | Collection[str] | None = None, seed: int | None = None, with_input_bias: bool = True, with_innovations: bool = True, with_lineage: bool = True)[source]¶

Create a structure handling CPPNs for generic use

Parameters:

inputs – Number of inputs for the CPPN
outputs – Number of outputs for the CPPN
labels – (Optional) list of i/o labels for pretty-printing
with_input_bias – Whether to use an input bias
seed – Seed for the random number generator
with_innovations – Whether to use NEAT historical markings
with_lineage – Whether to store lineage information

static create_for_eshn_cppn(dimension: int, with_input_bias: bool | None = True, with_input_length: bool | None = True, with_leo: bool | None = True, with_output_bias: bool | None = True, seed: int | None = None, with_innovations: bool = True, with_lineage: bool = True)[source]¶

Create a structure handling CPPNs for use with ES-HyperNEAT

Parameters:

dimension – The substrate dimension (2- or 3-D)
with_input_bias – Whether to use an input bias
with_input_length – Whether to directly provide the distance between points
with_leo – Whether to use a Level of Expression Output
with_output_bias – Whether to use the CPPN to generate per-neuron biases
seed – Seed for the random number generator
with_innovations – Whether to use NEAT historical markings
with_lineage – Whether to store lineage information

id() → int | None[source]¶: Return the genome id if one was generated

parents() → int | None[source]¶: Return the genome’s parent(s) if possible

classmethod random(data: Data) → Genome[source]¶

Create a random CPPN with boolean initialization

Parameters:: data – Evolution-wide shared genomic data
Returns:: A random CPPN genome

mutate(data: Data, n=1) → None[source]¶: Mutate (in-place) this genome :param data: Evolution-wide shared genomic data :param n: Number of mutations to perform

mutated(data: Data, n=1) → Genome[source]¶: Return a mutated (copied) version of this genome :param data: Evolution-wide shared genomic data :param n: Number of mutations to perform

static crossover(lhs: Genome, rhs: Genome, data: Data, bias: int | str = 'length') → Genome[source]¶

Create a new genome from two parents using historical markings

Bias specifies which parent to take disjoint and excess genes from:

0: lhs
1: rhs
“length”: from the smallest (or random if both have the same size)

Parameters:

lhs – A parent
rhs – Another parent
data – Evolution-wide shared genomic data
bias – Which parent to take disjoint and excess genes (see description)

Returns:

The child

static distance(lhs: Genome, rhs: Genome, weights=None) → float[source]¶

Compute the genetic distance between lhs and rhs

Parameters:

lhs – The first genome
rhs – The first genome
weights – How to weight the different terms:

Node excess
Node disjoint
Node matched
Link excess
Link disjoint
Link matched

Returns:: the weighted genetic distance

copy() → Genome[source]¶: Return a perfect (deep)copy of this genome

update_lineage(data: Data, parents: List[Genome])[source]¶

Update lineage fields

Parameters:

data – Evolution-wide shared genomic data
parents – list (potentially empty) of this genome’s parents

to_json() → Dict[Any, Any][source]¶: Return a json (dict) representation of this object

static from_json(data: Dict[Any, Any]) → Genome[source]¶: Recreate a Genome from a string json representation

static from_dot(path: str, rng: Random) → Genome[source]¶

Produce a Genome by parsing a simplified graph description

Warning

Unimplemented

Parameters:

path – the file to load
rng – random pick unspecified functions

Returns:

The user-specified CPPN

to_dot(data: Data, path: str, ext: str = 'pdf', title: str | None = None, debug: str | Iterable | None = None) → str[source]¶

Produce a graphical representation of this genome

Note

Missing functions images in nodes and/or lots of warning message are likely caused by misplaced image files. In doubt perform a full reinstall

Parameters:

data – Genome’s shared data (for labels)
path – The path to write to
ext – The rendering format to use
title – Optional title for the graph (e.g. for generational info)
debug –
Print more information on the graph. Special values (either as a single string or a list):
- ’depth’ will display every nodes’ depth
- ’links’ will display every links’ id
- ’keepdot’ will keep the intermediate dot file
- ’all’ all of the above

Raises:

OSError – if the dot program is not available (not installed, on the path and executable)

Helper types:¶

class abrain._cpp.genotype.Innovations¶

C++ database for innovation markings

NOT_FOUND = -1¶

copy(self: abrain._cpp.genotype.Innovations) → abrain._cpp.genotype.Innovations¶: Return a perfect (deep)copy of this innovations database

empty(self: abrain._cpp.genotype.Innovations) → bool¶: Whether any historical markings have been generated yet

static from_json(j: dict) → abrain._cpp.genotype.Innovations¶: Convert from the json-compliant Python dictionary j

get_link_id(self: abrain._cpp.genotype.Innovations, src: int, dst: int) → int¶

Attempt to generate a new link innovation marking for provided key

Parameters:

src – ID of the link’s source
dst – ID of the link’s destination

Returns:

The newly created, or existing corresponding link innovation marking

get_node_id(self: abrain._cpp.genotype.Innovations, src: int, dst: int) → int¶

Attempt to generate a new node innovation marking for provided key

Parameters:

src – ID of the link’s source
dst – ID of the link’s destination

Returns:

The newly created, or existing corresponding node innovation marking

initialize(self: abrain._cpp.genotype.Innovations, nextNodeID: int) → None¶

Reset the database.

Next node id is set to the provided value and the next link id is set to 0. All known nodes and links mappings are cleared.

Parameters:: nextNodeID – The next node id to give

link_id(self: abrain._cpp.genotype.Innovations, src: int, dst: int) → int¶

Retrieve link innovation marking for provided key

Parameters:

src – ID of the link’s source
dst – ID of the link’s destination

Returns:

The corresponding link innovation marking or Innovations.NOT_FOUND if not found

new_link_id(self: abrain._cpp.genotype.Innovations, src: int, dst: int) → int¶

Force generation of a new link innovation marking for provided key

Parameters:

src – ID of the link’s source
dst – ID of the link’s destination

Returns:

The newly created link innovation marking

new_node_id(self: abrain._cpp.genotype.Innovations, src: int, dst: int) → int¶

Force generation of a new node innovation marking for provided key

Parameters:

src – ID of the link’s source
dst – ID of the link’s destination

Returns:

The newly created node innovation marking

next_link_id(self: abrain._cpp.genotype.Innovations) → int¶: Historical marking of the next new link

next_node_id(self: abrain._cpp.genotype.Innovations) → int¶: Historical marking of the next new node

node_id(self: abrain._cpp.genotype.Innovations, src: int, dst: int) → int¶

Retrieve node innovation marking for provided key

The link refers to the link that was broken up to create this new: node

Parameters:

src – ID of the link’s source
dst – ID of the link’s destination

Returns:

The corresponding node innovation marking or Innovations.NOT_FOUND if not found

size(self: abrain._cpp.genotype.Innovations) → tuple¶: The number of node and link historical markings currently registered

to_json(self: abrain._cpp.genotype.Innovations) → dict¶: Convert to a json-compliant Python dictionary

class abrain.core.genome.IDManager[source]¶

Very simple monotonic ID generator for nodes and links

initialize(next_node_id: int)[source]¶: Reset the manager to next_node_id for the nodes and 0 for the links

get_node_id(*_)[source]¶

Request a new node id.

Same signature as get_node_id() but does not use the arguments and, instead, provides monotonously increasing ids

get_link_id(*_)[source]¶

Request a new link id.

Same signature as get_link_id() but does not use the arguments and, instead, provides monotonously increasing ids

next_node_id()[source]¶

The next id that would be assigned to a node.

Mostly for debugging purposes

next_link_id()[source]¶

The next id that would be assigned to a link.

Mostly for debugging purposes

link_id = None¶

node_id = None¶

new_node_id(*_)¶

Request a new node id.

Same signature as get_node_id() but does not use the arguments and, instead, provides monotonously increasing ids

new_link_id(*_)¶

Request a new link id.

Same signature as get_link_id() but does not use the arguments and, instead, provides monotonously increasing ids

class abrain.core.genome.GIDManager[source]¶: Simple integer-producing class for unique genome identifier

Underlying types:¶

class abrain._cpp.genotype.CPPNData¶

C++ supporting type for genomic data

class Link¶

From-to relationship between two computational node

dst¶: ID of the destination node

id¶: Historical marking

src¶: ID of the source node

weight¶: Connection weight

class Links¶

Collection of Links

append(self: abrain._cpp.genotype.CPPNData.Links, x: abrain._cpp.genotype.CPPNData.Link) → None¶: Add an item to the end of the list

clear(self: abrain._cpp.genotype.CPPNData.Links) → None¶: Clear the contents

extend(\*args, **kwargs)¶

Overloaded function.

extend(self: abrain._cpp.genotype.CPPNData.Links, L: abrain._cpp.genotype.CPPNData.Links) -> None

Extend the list by appending all the items in the given list

extend(self: abrain._cpp.genotype.CPPNData.Links, L: Iterable) -> None

Extend the list by appending all the items in the given list

insert(self: abrain._cpp.genotype.CPPNData.Links, i: int, x: abrain._cpp.genotype.CPPNData.Link) → None¶: Insert an item at a given position.

pop(\*args, **kwargs)¶

Overloaded function.

pop(self: abrain._cpp.genotype.CPPNData.Links) -> abrain._cpp.genotype.CPPNData.Link

Remove and return the last item

pop(self: abrain._cpp.genotype.CPPNData.Links, i: int) -> abrain._cpp.genotype.CPPNData.Link

Remove and return the item at index i

class Node¶

Computational node of a CPPN

func¶: Function used to compute

id¶: Historical marking

class Nodes¶

Collection of Nodes

append(self: abrain._cpp.genotype.CPPNData.Nodes, x: abrain._cpp.genotype.CPPNData.Node) → None¶: Add an item to the end of the list

clear(self: abrain._cpp.genotype.CPPNData.Nodes) → None¶: Clear the contents

extend(\*args, **kwargs)¶

Overloaded function.

extend(self: abrain._cpp.genotype.CPPNData.Nodes, L: abrain._cpp.genotype.CPPNData.Nodes) -> None

Extend the list by appending all the items in the given list

extend(self: abrain._cpp.genotype.CPPNData.Nodes, L: Iterable) -> None

Extend the list by appending all the items in the given list

insert(self: abrain._cpp.genotype.CPPNData.Nodes, i: int, x: abrain._cpp.genotype.CPPNData.Node) → None¶: Insert an item at a given position.

pop(\*args, **kwargs)¶

Overloaded function.

pop(self: abrain._cpp.genotype.CPPNData.Nodes) -> abrain._cpp.genotype.CPPNData.Node

Remove and return the last item

pop(self: abrain._cpp.genotype.CPPNData.Nodes, i: int) -> abrain._cpp.genotype.CPPNData.Node

Remove and return the item at index i

static from_json(j: dict) → abrain._cpp.genotype.CPPNData¶: Convert from the json-compliant Python dictionary j

to_json(self: abrain._cpp.genotype.CPPNData) → dict¶: Convert to a json-compliant Python dictionary

bias¶: Whether to use an input bias

inputs¶: Number of inputs

labels¶: (optional) label for the inputs/outputs

links¶: The collection of inter-node relationships

nodes¶: The collection of computing nodes

outputs¶: Number of outputs

Contains the classes and functions related to abrain’s genotype