Skip to main content
Home
Browse packagesDocsSign in
This release is 136 versions behind 0.168.1 — the latest version of @stsoftware/neat-ai. Jump to latest

@stsoftware/neat-ai@0.121.0
Built and signed on GitHub Actions

stSoftwareAU/NEAT-AI

NEAT Neural Network. This project is a unique implementation of a neural network based on the NEAT (NeuroEvolution of Augmenting Topologies) algorithm, written in DenoJS using TypeScript.

This package works with Deno
This package works with Deno
JSR Score
94%
Published
5 months ago (0.121.0)
Package root>test>TraceAggregate.ts
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170
import { assert, assertAlmostEquals } from "jsr:@std/assert@^1.0.8";
import { Creature } from "../src/Creature.ts";
import { createBackPropagationConfig } from "../src/propagate/BackPropagation.ts";
import type { CreatureInternal } from "../src/architecture/CreatureInterfaces.ts";
import { SparseConfig } from "../src/propagate/sparse/SparseConfig.ts";

((globalThis as unknown) as { DEBUG: boolean }).DEBUG = true;

Deno.test("TraceAggregateMINIMUM", () => {
  const json: CreatureInternal = {
    neurons: [
      { bias: 0.1, type: "hidden", squash: "LOGISTIC", index: 2 },
      { bias: -0.2, type: "hidden", squash: "LOGISTIC", index: 3 },
      { bias: 0.3, type: "hidden", squash: "MINIMUM", index: 4 },
      { bias: -0.4, type: "output", squash: "LOGISTIC", index: 5 },
      { bias: 0.5, type: "output", squash: "LOGISTIC", index: 6 },
    ],
    synapses: [
      { weight: 0.1, from: 0, to: 2 },
      { weight: -0.2, from: 1, to: 3 },
      { weight: 0.3, from: 2, to: 4 },
      { weight: -0.4, from: 3, to: 4 },
      { weight: -0.5, from: 4, to: 5 },
      { weight: 0.6, from: 4, to: 6 },
    ],
    input: 2,
    output: 2,
  };
  const creature = Creature.fromJSON(json);
  creature.validate();
  Deno.writeTextFileSync(
    "test/data/.a.json",
    JSON.stringify(creature.exportJSON(), null, 2),
  );
  const input = new Float32Array([0.1, 0.2]);
  creature.activate(input);

  const sparseConfig = new SparseConfig(
    creature.exportJSON(),
    createBackPropagationConfig({ sparseRatio: 1 }),
  );
  const aOut = creature.activateAndTrace(input, false, sparseConfig);

  const changed = creature.applyLearnings(
    createBackPropagationConfig({ trainingMutationRate: 1 }),
    sparseConfig,
  );

  assert(changed, "should have changed");

  const dOut = creature.activate(input);

  Deno.writeTextFileSync(
    "test/data/.d.json",
    JSON.stringify(creature.exportJSON(), null, 2),
  );
  assertAlmostEquals(aOut[0], dOut[0], 0.0001);

  assertAlmostEquals(aOut[1], dOut[1], 0.0001);
});

Deno.test("TraceAggregateMAXIMUM", () => {
  const json: CreatureInternal = {
    neurons: [
      { bias: 0.1, type: "hidden", squash: "LOGISTIC", index: 2 },
      { bias: -0.2, type: "hidden", squash: "LOGISTIC", index: 3 },
      { bias: 0.3, type: "hidden", squash: "MAXIMUM", index: 4 },
      { bias: -0.4, type: "output", squash: "LOGISTIC", index: 5 },
      { bias: 0.5, type: "output", squash: "LOGISTIC", index: 6 },
    ],
    synapses: [
      { weight: 0.1, from: 0, to: 2 },
      { weight: -0.2, from: 1, to: 3 },
      { weight: 0.3, from: 2, to: 4 },
      { weight: -0.4, from: 3, to: 4 },
      { weight: -0.5, from: 4, to: 5 },
      { weight: 0.6, from: 4, to: 6 },
    ],
    input: 2,
    output: 2,
  };
  const creature = Creature.fromJSON(json);
  creature.validate();
  Deno.writeTextFileSync(
    "test/data/.A.json",
    JSON.stringify(creature.exportJSON(), null, 2),
  );
  const input = new Float32Array([0.1, 0.2]);
  creature.activate(input);

  const sparseConfig = new SparseConfig(
    creature.exportJSON(),
    createBackPropagationConfig({
      sparseRatio: 1,
    }),
  );
  const aOut = creature.activateAndTrace(input, false, sparseConfig);

  const changed = creature.applyLearnings(
    createBackPropagationConfig({ trainingMutationRate: 1 }),
    sparseConfig,
  );

  assert(changed, "should have changed");

  const dOut = creature.activate(input);

  Deno.writeTextFileSync(
    "test/data/.B.json",
    JSON.stringify(creature.exportJSON(), null, 2),
  );
  assertAlmostEquals(aOut[0], dOut[0], 0.0001);

  assertAlmostEquals(aOut[1], dOut[1], 0.0001);
});

Deno.test("TraceAggregateIF", () => {
  const json: CreatureInternal = {
    neurons: [
      { bias: 0.1, type: "hidden", squash: "LOGISTIC", index: 2 },
      { bias: -0.2, type: "hidden", squash: "LOGISTIC", index: 3 },
      { bias: 0.3, type: "hidden", squash: "IF", index: 4 },
      { bias: -0.4, type: "output", squash: "LOGISTIC", index: 5 },
      { bias: 0.5, type: "output", squash: "LOGISTIC", index: 6 },
    ],
    synapses: [
      { weight: 0.1, from: 0, to: 2 },
      { weight: -0.2, from: 1, to: 3 },
      { weight: 0.15, from: 1, to: 4, type: "condition" },
      { weight: 0.3, from: 2, to: 4, type: "positive" },
      { weight: -0.4, from: 3, to: 4, type: "negative" },
      { weight: -0.5, from: 4, to: 5 },
      { weight: 0.6, from: 4, to: 6 },
    ],
    input: 2,
    output: 2,
  };
  const creature = Creature.fromJSON(json);
  creature.validate();
  Deno.writeTextFileSync(
    "test/data/.a.json",
    JSON.stringify(creature.exportJSON(), null, 2),
  );
  const input = new Float32Array([0.1, 0.2]);
  creature.activate(input);

  const sparseConfig = new SparseConfig(
    creature.exportJSON(),
    createBackPropagationConfig({ sparseRatio: 1 }),
  );

  const aOut = creature.activateAndTrace(input, false, sparseConfig);

  const changed = creature.applyLearnings(
    createBackPropagationConfig({ trainingMutationRate: 1 }),
    sparseConfig,
  );

  assert(changed, "should have changed");

  const dOut = creature.activate(input);

  Deno.writeTextFileSync(
    "test/data/.d.json",
    JSON.stringify(creature.exportJSON(), null, 2),
  );
  assertAlmostEquals(aOut[0], dOut[0], 0.0001);

  assertAlmostEquals(aOut[1], dOut[1], 0.0001);
});