Module gp_cgp 

Cartesian Genetic Programming.

CGP encodes a directed acyclic computation graph on a fixed rows × cols grid. Each node stores (function_id, input_0, input_1), plus the final output gene picks which node produces the output. The genotype is a fixed-length integer vector, so populations are Tensor<B, 2, Int> and fit the tensor abstraction cleanly.

Evolutionary engine

Canonical CGP uses a (1 + λ) Evolution Strategy with point mutation and no crossover. This module re-implements just that engine directly — not via crate::algorithms::es_classical — so the mutation logic can be specialized to the CGP genome semantics (constrained feed-forward connections, function_id range, …).

Function set

The v1 function set is fixed at construction time:

id	op	arity	formula
0	add	2	a + b
1	sub	2	a − b
2	mul	2	a · b
3	protected_div	2	a / b (or a if `
4	sin	1	sin(a)
5	cos	1	cos(a)
6	tanh	1	tanh(a)
7	const 1.0	0	1.0

Phenotype evaluation

Evaluation runs on the host because the per-node dispatch is not a good fit for dense tensor ops; node values are computed in topological order (left-to-right across the grid columns). Genotype storage stays on-device to match the other strategies.

Reference

• Miller (2011), Cartesian Genetic Programming (Natural Computing Series).

Structs

CartesianGeneticProgramming

Classical Cartesian GP with (1 + λ) ES.

CgpConfig

Static configuration for a CartesianGeneticProgramming run.

CgpState

Generation state for CartesianGeneticProgramming.

Constants

FUNCTION_ARITIES

Fixed v1 function set: arity of each opcode.

NUM_FUNCTIONS

Number of opcodes in the v1 function set.

Functions

evaluate_cgp

Evaluates a CGP genotype at a set of input rows.