reduce grid by factor 10, to allow for Discrete to be more visible in

effect in examples/visualize_evolve_mutation_types and examples/visualize_permutate_mutation_types

Author: basvanwesting

README.md

@@ -140,7 +140,7 @@ within a single numeric type `T`.
 
 The library supports various mutation strategies that affect how the genetic
 algorithm explores the search space. Random leads to the best results overall.
-Random is the default and is supported by all Genotypes. 
+Random is the default and is supported by all Genotypes.
 
 But for numeric genotypes (RangeGenotype and MultiRangeGenotype) there are
 several alternatives. These might converge faster, but are all more sensitive to
@@ -163,9 +163,12 @@ Run the example with `cargo run --example visualize_evolve_mutation_types --rele
 
 ### Permutate Strategy
 
-For exhaustive search in smaller spaces, the Permutate strategy can
-systematically explore continues genotypes (RangeGenotype and
-MultiRangeGenotype) using Step, StepScaled, and Discrete mutation types:
+For exhaustive search in smaller spaces, all genotypes have their own
+permutation implementation, which systematically explore all value combinations.
+
+But for numeric/continues genotypes (RangeGenotype and MultiRangeGenotype)
+permutation is only possible using Step, StepScaled, and Discrete mutation
+types (as it needs additional restrictions be become countable):
 
 ![Permutate Mutation Types Patterns](examples/visualize_permutate_mutation_types.png)
 

examples/visualize_evolve_mutation_types.png

@@ -3,7 +3,7 @@ use plotters::prelude::*;
 use std::sync::{Arc, Mutex};
 
 const SEED_POINT: [f32; 2] = [0.0, 0.0];
-const TARGET_POINT: [f32; 2] = [66.666, 77.777];
+const TARGET_POINT: [f32; 2] = [6.6666, 7.7777];
 
 /// Fitness function targeting the point in 2D space
 #[derive(Clone, Debug)]
@@ -121,13 +121,13 @@ fn run_evolution(
     mutation_type_name: String,
     max_stale_generations: usize,
 ) -> ExplorationReporter {
-    let fitness = TargetPointFitness::new(TARGET_POINT, 0.001);
+    let fitness = TargetPointFitness::new(TARGET_POINT, 0.0001);
     let reporter = ExplorationReporter::new();
 
     // Create genotype with 2 genes for 2D visualization
     let genotype = RangeGenotype::<f32>::builder()
         .with_genes_size(2)
-        .with_allele_range(0.0..=100.0)
+        .with_allele_range(0.0..=10.0)
         .with_mutation_type(mutation_type)
         .with_seed_genes_list(vec![SEED_POINT.to_vec()]) // Start from origin
         .build()
@@ -182,7 +182,7 @@ fn generate_plot(
             .margin(10)
             .x_label_area_size(30)
             .y_label_area_size(30)
-            .build_cartesian_2d(0f32..100f32, 0f32..100f32)?;
+            .build_cartesian_2d(0f32..10f32, 0f32..10f32)?;
 
         chart.configure_mesh().draw()?;
 
@@ -271,7 +271,7 @@ fn main() {
     println!("=== Visualizing Evolve Mutation Types in 2D Search Space ===\n");
     println!("Starting point: {:?}", SEED_POINT);
     println!("Target point: {:?}", TARGET_POINT);
-    println!("Search space: [0.0, 100.0] x [0.0, 100.0]\n");
+    println!("Search space: [0.0, 10.0] x [0.0, 10.0]\n");
 
     let mut reporters = Vec::new();
 
@@ -281,13 +281,13 @@ fn main() {
     reporters.push(("Random".to_string(), reporter));
 
     // Range mutation - fixed bandwidth
-    println!("Running Range(10.0) mutation...");
-    let reporter = run_evolution(MutationType::Range(10.0), "Range(10.0)".to_string(), 20);
-    reporters.push(("Range(±10)".to_string(), reporter));
+    println!("Running Range(1.0) mutation...");
+    let reporter = run_evolution(MutationType::Range(1.0), "Range(1.0)".to_string(), 20);
+    reporters.push(("Range(±1.0)".to_string(), reporter));
 
     // RangeScaled mutation - dropping by 1/5th each scale
     println!("Running RangeScaled mutation...");
-    let ranges = vec![100.0, 100.0, 100.0, 90.0, 30.0, 1.0];
+    let ranges = vec![10.0, 10.0, 10.0, 9.0, 3.0, 0.1];
     let reporter = run_evolution(
         MutationType::RangeScaled(ranges.clone()),
         format!("RangeScaled({:?})", ranges),
@@ -296,13 +296,13 @@ fn main() {
     reporters.push(("RangeScaled (slow starting sigmoid)".to_string(), reporter));
 
     // Step mutation - fixed step
-    println!("Running Step(1.0) mutation...");
-    let reporter = run_evolution(MutationType::Step(1.0), "Step(1.0)".to_string(), 20);
-    reporters.push(("Step(±1)".to_string(), reporter));
+    println!("Running Step(0.3) mutation...");
+    let reporter = run_evolution(MutationType::Step(0.3), "Step(0.3)".to_string(), 20);
+    reporters.push(("Step(±0.3)".to_string(), reporter));
 
     // StepScaled mutation - halving each scale
     println!("Running StepScaled mutation...");
-    let steps = vec![50.0, 25.0, 12.5, 6.25, 3.125, 1.5625];
+    let steps = vec![5.00, 2.50, 1.25, 0.625, 0.3125, 0.15625];
     let reporter = run_evolution(
         MutationType::StepScaled(steps.clone()),
         format!("StepScaled({:?})", steps),
@@ -314,7 +314,7 @@ fn main() {
     println!("Running Discrete mutation...");
     let reporter = run_evolution(MutationType::Discrete, "Discrete".to_string(), 50);
     reporters.push((
-        "Discrete (floored integers, map to categories)".to_string(),
+        "Discrete (integers, map to categories)".to_string(),
         reporter,
     ));
 

examples/visualize_evolve_mutation_types.rs

@@ -2,7 +2,7 @@ use genetic_algorithm::strategy::permutate::prelude::*;
 use plotters::prelude::*;
 use std::sync::{Arc, Mutex};
 
-const TARGET_POINT: [f32; 2] = [66.666, 77.777];
+const TARGET_POINT: [f32; 2] = [6.6666, 7.7777];
 
 /// Fitness function targeting the point in 2D space
 #[derive(Clone, Debug)]
@@ -88,14 +88,14 @@ fn run_permutation(
     mutation_type: MutationType<f32>,
     mutation_type_name: String,
 ) -> PermutationReporter {
-    let fitness = TargetPointFitness::new(TARGET_POINT, 0.001);
+    let fitness = TargetPointFitness::new(TARGET_POINT, 0.0001);
     let reporter = PermutationReporter::new();
 
     // Create genotype with 2 genes for 2D visualization
     // Using integer type for discrete steps
     let genotype = RangeGenotype::<f32>::builder()
         .with_genes_size(2)
-        .with_allele_range(0.0..=100.0)
+        .with_allele_range(0.0..=10.0)
         .with_mutation_type(mutation_type)
         // No seed - let permutation explore the entire space
         .build()
@@ -143,7 +143,7 @@ fn generate_plot(
             .margin(10)
             .x_label_area_size(30)
             .y_label_area_size(30)
-            .build_cartesian_2d(0f32..100f32, 0f32..100f32)?;
+            .build_cartesian_2d(0f32..10f32, 0f32..10f32)?;
 
         chart.configure_mesh().draw()?;
 
@@ -220,32 +220,34 @@ fn generate_plot(
 fn main() {
     println!("=== Visualizing Permutation Mutation Types in 2D Search Space ===\n");
     println!("Target point: {:?}", TARGET_POINT);
-    println!("Search space: [0, 100] x [0, 100] (integer grid)\n");
-    println!("Note: Permutation systematically explores the entire space as defined by the mutation type.\n");
+    println!("Search space: [0.0, 10.0] x [0.0, 10.0]\n");
 
     let mut reporters = Vec::new();
 
     // Step mutation - fixed step
     // With step=5, this will explore a 20x20 grid (every 5th point)
-    println!("Running Step(5) mutation with permutation...");
-    let reporter = run_permutation(MutationType::Step(5.0), "Step(5.0)".to_string());
-    reporters.push(("Step(±5) Full Grid Exploration".to_string(), reporter));
+    println!("Running Step(0.3) mutation...");
+    let reporter = run_permutation(MutationType::Step(0.3), "Step(0.3)".to_string());
+    reporters.push(("Step(±0.3, full exploration)".to_string(), reporter));
 
     // StepScaled mutation - halving each scale
     // This will first explore with large steps, then refine around the best
-    println!("Running StepScaled mutation with permutation...");
-    let steps = vec![50.0, 25.0, 12.5, 6.25, 3.125, 1.5625];
+    println!("Running StepScaled mutation...");
+    let steps = vec![5.00, 2.50, 1.25, 0.625, 0.3125, 0.15625];
     let reporter = run_permutation(
         MutationType::StepScaled(steps.clone()),
         format!("StepScaled({:?})", steps),
     );
-    reporters.push(("StepScaled Nested Grids".to_string(), reporter));
+    reporters.push(("StepScaled (nested halving grids)".to_string(), reporter));
 
     // Discrete mutation - like ListGenotype for categories
     // This explores a sampled subset of the space
-    println!("Running Discrete mutation with permutation (limited range for feasibility)...");
+    println!("Running Discrete mutation...");
     let reporter = run_permutation(MutationType::Discrete, "Discrete".to_string());
-    reporters.push(("Discrete Full Exploration".to_string(), reporter));
+    reporters.push((
+        "Discrete (full exploration of integers)".to_string(),
+        reporter,
+    ));
 
     // Generate visualization
     println!("\nGenerating visualization...");
@@ -259,4 +261,3 @@ fn main() {
     println!("- Discrete: Jumps to any discrete value in the range");
     println!("Color gradient: Blue (early) → Green (late) generations");
 }
-

examples/visualize_permutate_mutation_types.png

@@ -143,9 +143,12 @@
 //!
 //! ### Permutate Strategy (and HillClimb)
 //!
-//! For exhaustive search in smaller spaces, the Permutate strategy can
-//! systematically explore continues genotypes (RangeGenotype and
-//! MultiRangeGenotype) using Step, StepScaled, and Discrete mutation types:
+//! For exhaustive search in smaller spaces, all genotypes have their own
+//! permutation implementation, which systematically explore all value combinations.
+//!
+//! But for numeric/continues genotypes (RangeGenotype and MultiRangeGenotype)
+//! permutation is only possible using Step, StepScaled, and Discrete mutation
+//! types (as it needs additional restrictions be become countable):
 //!
 //! - **Step**: Systematically explores grid points at fixed intervals
 //! - **StepScaled**: Hierarchical search that refines around promising regions