add slight bias to parent selection

README.md

@@ -1,3 +1,7 @@
 Golang implementation of the GENTIOR TSP heuristic  developed ad Colorado State University with D. Fuquay and D. Whitley
 
 The GENTIOR is described in Fuquay and Whitley's 1990 paper [Genetic Algorithm Solutions for the Traveling Salesman Problem](https://dl.acm.org/citation.cfm?id=99033&dl=ACM&coll=DL)
+
+TODO:
+- Add some biased parent selection methods to compare convergence
+- Inspect for complexity issues

main.go

@@ -100,19 +100,20 @@ func printProgress(i, n int) {
 	fmt.Printf("%d%% \n", int(percent))
 }
 
+
 func ConnectedGentior(filepath string, populationSize, generations int, bias float64, quiet bool) {
     start := time.Now()
 	// Create graph and population
-	graph := NewWeightedGraphFromFile(filepath)
-	population := generatepopulation(graph, populationSize)
+	graph := NewWeightedGraphFromFile(filepath) //O(n)
+	population := generatepopulation(graph, populationSize) //O(p)
 
 	// genetically develop good solutions (hopefully)
-	for i := 0; i < generations; i++ {
-		parents := selectParents(len(population), bias)
-		pop := population[parents[0]]
-		mom := population[parents[1]]
-		child := connectedEdgeRecombination(pop, mom, graph)
-		population = reconstructPopulation(population, child)
+	for i := 0; i < generations; i++ { // O(n)
+		parents := selectParents(len(population), bias)         // O(1)
+		pop := population[parents[0]]                           // O(1)
+		mom := population[parents[1]]                           // O(1)
+		child := connectedEdgeRecombination(pop, mom, graph)    // O(n)
+		population = reconstructPopulation(population, child)   // O(logn)
 		if !quiet {
 			printProgress(i, generations)
 		}
@@ -393,20 +394,52 @@ func showPopulation(population []Hamiltonian) {
 }
 
 func selectParents(populationSize int, bias float64) []int {
+	return betterParents(populationSize, bias)
+}
+
+func randomParents(populationSize int, bias float64) []int {
 	// Select at Random
 	// TODO implement bias
 	mom := rand.Intn(populationSize)
 	pop := rand.Intn(populationSize)
 
 	// ensure parents are different
 	for mom == pop {
-		pop = rand.Intn(populationSize)
+	pop = rand.Intn(populationSize)
+	}
+
+	parentList := []int{pop, mom}
+	return parentList
+}
+
+func betterParents(populationSize int, bias float64) []int {
+	// Select at Random
+	// TODO implement bias
+	mom := betterRand(populationSize, bias)
+	pop := betterRand(populationSize, bias)
+
+	// ensure parents are different
+	for mom == pop {
+	pop = betterRand(populationSize, bias)
 	}
 
 	parentList := []int{pop, mom}
 	return parentList
 }
 
+func betterRand(max int, bias float64) int {
+    // chose two indices
+    n1 := rand.Intn(max)
+    n2 := rand.Intn(max)
+
+    // return the lesser of the two
+    if n1 < n2 {
+        return n1
+    } else {
+        return n2
+    }
+}
+
 //TODO?
 func applyBias(i int, max int, b float64) int {
 	//probabilities := make(

data/XQF131/tsp2wdat.py → tools/tsp2wdat.py

@@ -1,11 +1,17 @@
 import math
 
-outFilename = "./xqf131.wdat"
-outFile = open(outFilename, "w")
+# tsp2wdat transforms a .tsp file into a .wdat (weighted data) file.
+# A .wdat has a header: the number of vertices in the graph, followed by a
+# number of lines descriving each edge. Edges are described by the vertex numbers
+# of each end of the edge, followed by the distance spanned by the edge.
+# Finally, the EOF indicator is the footer "-1 -1 -1"
 
 filename = "./xqf131.tsp"
 edgeFile = open(filename)
 
+outFilename = "./xqf131.wdat"
+outFile = open(outFilename, "w")
+
 string = ""
 
 numPoints = 0