Implemented integration test for double-pole balancing experiment wit…

…h parallel fitness evaluation.

examples/pole2/cart2pole_parallel_test.go

@@ -0,0 +1,86 @@
+package pole2
+
+import (
+	"fmt"
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+	"github.com/yaricom/goNEAT/v4/examples/utils"
+	"github.com/yaricom/goNEAT/v4/experiment"
+	"github.com/yaricom/goNEAT/v4/neat"
+	"math/rand"
+	"testing"
+)
+
+func TestCartDoublePoleParallelGenerationEvaluator_GenerationEvaluate_Markov(t *testing.T) {
+	if testing.Short() {
+		t.Skip("skipping test in short Unit Test mode.")
+	}
+
+	// to make sure we have predictable results
+	rand.Seed(423)
+
+	outDirPath, contextPath, genomePath := "../../out/pole2_markov_parallel_test", "../../data/pole2_markov.neat", "../../data/pole2_markov_startgenes"
+
+	fmt.Println("Loading start genome for POLE2 Markov parallel experiment")
+	// Load NEAT options and initial genome
+	opts, startGenome, err := utils.LoadOptionsAndGenome(contextPath, genomePath)
+	neat.LogLevel = neat.LogLevelInfo
+	require.NoError(t, err)
+
+	// Check if output dir exists
+	err = utils.CreateOutputDir(outDirPath)
+	require.NoError(t, err, "Failed to create output directory")
+
+	// Running POLE2 Markov parallel experiment
+	opts.NumRuns = 5
+	exp := experiment.Experiment{
+		Id:     0,
+		Trials: make(experiment.Trials, opts.NumRuns),
+	}
+	evaluator := NewCartDoublePoleParallelGenerationEvaluator(outDirPath, true, ContinuousAction)
+	err = exp.Execute(opts.NeatContext(), startGenome, evaluator, nil)
+	require.NoError(t, err, "Failed to perform POLE2 Markov parallel experiment")
+
+	// Find winner statistics
+	avgNodes, avgGenes, avgEvals, _ := exp.AvgWinnerStatistics()
+
+	// check results
+	if avgNodes < 8 {
+		t.Error("avg_nodes < 8", avgNodes)
+	} else if avgNodes > 40 {
+		t.Error("avg_nodes > 40", avgNodes)
+	}
+
+	if avgGenes < 7 {
+		t.Error("avg_genes < 7", avgGenes)
+	} else if avgGenes > 50 {
+		t.Error("avg_genes > 50", avgGenes)
+	}
+
+	maxEvals := float64(opts.PopSize * opts.NumGenerations)
+	assert.True(t, avgEvals < maxEvals)
+
+	t.Logf("Average nodes: %.1f, genes: %.1f, evals: %.1f\n", avgNodes, avgGenes, avgEvals)
+	meanComplexity, meanDiversity, meanAge := 0.0, 0.0, 0.0
+	for _, t := range exp.Trials {
+		meanComplexity += t.ChampionsComplexities().Mean()
+		meanDiversity += t.Diversity().Mean()
+		meanAge += t.ChampionSpeciesAges().Mean()
+	}
+	count := float64(len(exp.Trials))
+	meanComplexity /= count
+	meanDiversity /= count
+	meanAge /= count
+	t.Logf("Mean best organisms: complexity=%.1f, diversity=%.1f, age=%.1f\n", meanComplexity, meanDiversity, meanAge)
+
+	solvedTrials := 0
+	for _, tr := range exp.Trials {
+		if tr.Solved() {
+			solvedTrials++
+		}
+	}
+
+	t.Logf("Trials solved/run: %d/%d", solvedTrials, len(exp.Trials))
+
+	assert.NotZero(t, solvedTrials, "Failed to solve at least one trial. Need to be checked what was going wrong")
+}

examples/pole2/cart2pole_test.go

@@ -12,7 +12,7 @@ import (
 )
 
 // Run double pole-balancing experiment with Markov environment setup
-func TestCartDoublePoleGenerationEvaluator_GenerationEvaluateMarkov(t *testing.T) {
+func TestCartDoublePoleGenerationEvaluator_GenerationEvaluate_Markov(t *testing.T) {
 	if testing.Short() {
 		t.Skip("skipping test in short Unit Test mode.")
 	}
@@ -87,7 +87,7 @@ func TestCartDoublePoleGenerationEvaluator_GenerationEvaluateMarkov(t *testing.T
 }
 
 // Run double pole-balancing experiment with Non-Markov environment setup
-func TestCartDoublePoleGenerationEvaluator_GenerationEvaluateNonMarkov(t *testing.T) {
+func TestCartDoublePoleGenerationEvaluator_GenerationEvaluate_NonMarkov(t *testing.T) {
 	if testing.Short() {
 		t.Skip("skipping test in short Unit Test mode.")
 	}