genetically_evolving_neural_network 0.1.0

This package implements a Genetically Evolving Neural Network. A Population of Entities, each containing a mutable Neural Network to drive decisions, will reproduce to create a new Generation (through [...]

example/example.md

Logical XOR Example

This example will attempt to correctly guess if a set of inputs satisfies Logical XOR - meaning one and only one input should be True (represented by 1.0), and all other inputs should be False (represented by 0.0).

You can access the full example by cloning this package locally, here.

logical_xor_fitness_service.dart

This file contains your custom FitnessService that will reward correctly guessing if a set of inputs satisfies Logical XOR.

import 'dart:math';
import 'package:genetically_evolving_neural_network/genetically_evolving_neural_network.dart';

/// This fitness service will be used to score a logical XOR calculator. The
/// output should only return true if a single value is 1.0 and both other
/// values are 0.0. There should be one exclusive positive value! The more
/// correct guesses that a NeuralNetwork makes, the higher its fitness score
/// will be.
class LogicalXORFitnessService extends GENNFitnessService {
  /// The list of logical inputs for your XOR calculator. The possible options
  /// are 0 or 1, effectively true or false.
  List<List<double>> logicalInputsList = [
    [0.0, 0.0, 0.0],
    [0.0, 0.0, 1.0],
    [0.0, 1.0, 0.0],
    [0.0, 1.0, 1.0],
    [1.0, 0.0, 0.0],
    [1.0, 0.0, 1.0],
    [1.0, 1.0, 0.0],
    [1.0, 1.0, 1.0],
  ];

  /// The list of logical outputs for your XOR calculator. These are the
  /// expected outputs respective to the logical inputs.
  List<List<double>> targetOutputsList = [
    [0.0],
    [1.0],
    [1.0],
    [0.0],
    [1.0],
    [0.0],
    [0.0],
    [0.0],
  ];

  /// This function will calculate a fitness score after guessing with every
  /// input within [LogicalXORFitnessService.logicalInputsList] on the input
  /// [neuralNetwork].
  @override
  Future<double> gennScoringFunction({
    required GENNNeuralNetwork neuralNetwork,
  }) async {
    // Collect all the guesses from this NeuralNetwork
    final guesses = getNeuralNetworkGuesses(neuralNetwork: neuralNetwork);

    // Declare a variable to store the sum of all errors
    var errorSum = 0.0;

    // Cycle through each guess to check its validity
    for (int i = 0; i < guesses.length; i++) {
      // Calculate the error from this guess
      final error = (targetOutputsList[i][0] - guesses[i][0]).abs();

      // Add this error to the errorSum
      errorSum += error;
    }

    // Calculate the difference between a perfect score (8) and the total
    // errors. A perfect score would mean zero errors with 8 correct answers,
    // meaning a perfect score would be 8.
    final diff = logicalInputsList.length - errorSum;

    // To make the better performing Entities stand out more in this population,
    // use the following equation to calculate the FitnessScore.
    //
    // 4 to the power of diff
    return pow(4, diff).toDouble();
  }

  /// Returns the list of guesses (or outputs) from the input [neuralNetwork]
  /// based on the standard set of inputs,
  /// [LogicalXORFitnessService.logicalInputsList].
  List<List<double>> getNeuralNetworkGuesses({
    required GENNNeuralNetwork neuralNetwork,
  }) {
    // Declare a list of guesses
    List<List<double>> guesses = [];

    // Cycle through each input
    for (int i = 0; i < logicalInputsList.length; i++) {
      // Declare this run's set of inputs
      final inputs = logicalInputsList[i];

      // Make a guess using the NeuralNetwork
      final guess = neuralNetwork.guess(inputs: inputs);

      // Add this guess to the list of guesses
      guesses.add(guess);
    }

    // Return the list of guesses
    return guesses;
  }
}

main.dart

This file contains the logic around setting up this visual example.

import 'dart:math';
import 'package:flutter/material.dart';
import 'package:genetically_evolving_neural_network/genetically_evolving_neural_network.dart';
import 'package:logical_xor/diagram_key.dart';
import 'package:logical_xor/logical_xor_fitness_service.dart';
import 'package:logical_xor/ui_helper.dart';

void main() {
  runApp(const MyApp());
}

class MyApp extends StatefulWidget {
  const MyApp({
    super.key,
  });

  @override
  State<MyApp> createState() => _MyAppState();
}

class _MyAppState extends State<MyApp> {
  /// Whether the example is currently playing forward.
  bool isPlaying = false;

  /// The highest possible fitness score.
  final double targetFitnessScore =
      pow(4, 8) + LogicalXORFitnessService().nonZeroBias;

  /// The Genetically Evolving Neural Network object.
  late final GENN genn;

  /// The current generation of Neural Networks.
  GENNGeneration? generation;

  /// The first wave to contain an Entity that reached the target fitness score.
  int? waveTargetFound;

  /// The number of inputs being fed into the Neural Network.
  static const numInitialInputs = 3;

  /// This represents the time to wait between waves shown on screen during
  /// continuous play.
  static const waitTimeBetweenWaves = 400;

  /// Whether to continuously play on through generations, or only increment by
  /// a single generation after each "play" click.
  bool continuousPlay = true;

  /// Used to build components of this example file's UI that are not related to
  /// understanding how the GENN class works.
  static UIHelper get uiHelper => UIHelper(numInitialInputs: numInitialInputs);
  @override
  void initState() {
    // Declare a config with specific mutation rates.
    final config = GENNGeneticEvolutionConfig(
      populationSize: 40,
      numOutputs: 1,
      mutationRate: 0.05,
      numInitialInputs: numInitialInputs,
      layerMutationRate: 0.10,
      perceptronMutationRate: 0.2,
      trackParents: true,
      // We only care about tracking the parents of the current generation to
      // show on-screen
      generationsToTrack: 1,
    );

    // Create your Genetically Evolving Neural Network object.
    genn = GENN.create(
      config: config,
      fitnessService: LogicalXORFitnessService(),
    );

    // Initialize the first generation
    genn.nextGeneration().then((value) {
      setState(() {
        generation = value;
      });
    });

    super.initState();
  }

  @override
  Widget build(BuildContext context) {
    // Necessary for initial loading of the screen.
    final generation = this.generation;
    if (generation == null) {
      return const CircularProgressIndicator();
    }

    WidgetsBinding.instance.addPostFrameCallback((timeStamp) async {
      if (isPlaying) {
        // Sleep for [waitTimeBetweenWaves] during continuous play so that the
        // gradual evolution changes are easier to see.
        await Future.delayed(
            const Duration(milliseconds: waitTimeBetweenWaves));

        // Create and set the next Generation to be displayed
        genn.nextGeneration().then((value) {
          setState(() {
            this.generation = value;
          });
        });
      }
    });

    // Check if target has been found.
    if (waveTargetFound == null &&
        generation.population.topScoringEntity.fitnessScore ==
            targetFitnessScore) {
      waveTargetFound = generation.wave;
    }

    final topScoringParents = generation.population.topScoringEntity.parents;
    return MaterialApp(
      title: 'Flutter Demo',
      theme: ThemeData(
        colorScheme: ColorScheme.fromSeed(seedColor: Colors.deepPurple),
        useMaterial3: true,
      ),
      home: Scaffold(
        body: SafeArea(
          child: Row(
            children: [
              const DiagramKey(numInitialInputs: numInitialInputs),
              SizedBox(
                width: MediaQuery.of(context).size.width / 2,
                child: Column(
                  crossAxisAlignment: CrossAxisAlignment.center,
                  children: [
                    const SizedBox(height: 24.0),
                    if (topScoringParents != null)
                      const Text(
                        'Parents of Top Performing Neural Network',
                        style: TextStyle(fontWeight: FontWeight.bold),
                      ),
                    if (topScoringParents != null)
                      Column(
                        children: [
                          uiHelper.showPerceptronMapWithScore(
                              entity: topScoringParents[0]),
                          uiHelper.showPerceptronMapWithScore(
                              entity: topScoringParents[1]),
                        ],
                      ),
                    const SizedBox(height: 12.0),
                    const Text(
                      ' Top Performing Neural Network',
                      style: TextStyle(fontWeight: FontWeight.bold),
                    ),
                    Text(
                      'Generation: ${generation.wave.toString()}',
                    ),
                    uiHelper.showPerceptronMapWithScore(
                      entity: generation.population.topScoringEntity,
                      showLabels: true,
                    ),
                    Text(
                      'Target Score: $targetFitnessScore',
                    ),
                    Row(
                      mainAxisAlignment: MainAxisAlignment.center,
                      children: [
                        Column(
                          children: [
                            Row(
                              children: [
                                uiHelper.showLogicalInputs(),
                                const SizedBox(width: 12),
                                uiHelper.showCorrectAnswers(),
                                const SizedBox(width: 12),
                                uiHelper.showNeuralNetworkGuesses(
                                  generation.population.topScoringEntity,
                                ),
                              ],
                            ),
                          ],
                        ),
                      ],
                    ),
                    if (waveTargetFound != null)
                      Text('Target reached at Generation: $waveTargetFound'),
                    const SizedBox(height: 24),
                    Text(
                      'Entire Population of Neural Networks (${generation.population.entities.length} in total)',
                    ),
                    const Text(
                      'These are chosen as parents to breed the next generation',
                      style: TextStyle(fontStyle: FontStyle.italic),
                    ),
                    uiHelper.perceptronMapDivider,
                    Flexible(
                      child: ListView.separated(
                        itemBuilder: (context, index) =>
                            uiHelper.showPerceptronMapWithScore(
                          entity: generation.population.sortedEntities[index],
                        ),
                        itemCount: generation.population.entities.length,
                        // itemCount: 16,
                        separatorBuilder: (context, index) =>
                            uiHelper.perceptronMapDivider,
                      ),
                    ),
                  ],
                ),
              ),
            ],
          ),
        ),
        floatingActionButtonLocation: FloatingActionButtonLocation.endContained,
        floatingActionButton: Column(
          mainAxisAlignment: MainAxisAlignment.end,
          children: [
            Text('ContinuousPlay: ${continuousPlay ? 'On' : 'Off'}'),
            Switch.adaptive(
                value: continuousPlay,
                onChanged: (value) {
                  setState(() {
                    continuousPlay = value;
                    isPlaying = false;
                  });
                }),
            FloatingActionButton(
              onPressed: () {
                if (continuousPlay) {
                  setState(() {
                    isPlaying = !isPlaying;
                  });
                } else {
                  genn.nextGeneration().then((value) {
                    setState(() {
                      this.generation = value;
                    });
                  });
                }
              },
              child: (!continuousPlay || !isPlaying)
                  ? const Icon(Icons.play_arrow)
                  : const Icon(Icons.pause),
            ),
            const SizedBox(height: 12.0),
          ],
        ),
      ),
    );
  }
}

perceptron_map

This directory contains many files, and would bloat this example page further.

You can see the PerceptronMap directory here.