go: collect and process mean vals

* compute and save mean values of x (e.g. 30) bench runs
* safely save mean vals for all bench-dimens combinations to AlgoMeans
  from goroutines (use mutex to protect shared access)
* rework algo/plot.go to use already computed mean values.

todo: plot mean vals of all algorithms (per benchmark) in a single pic.

algo/algo.go

@@ -15,6 +15,11 @@ import (
 // methods over it.
 type Values []float64
 
+// mu protects access to meanStats.
+var mu sync.Mutex
+
+var meanStats = &stats.MeanStats{}
+
 // DoRandomSearch executes a search using the 'Random search' method.
 func DoRandomSearch(wg *sync.WaitGroup, m *sync.Mutex) {
 	defer wg.Done()

algo/plot.go

@@ -21,7 +21,7 @@ import (
 
 const preferredFontStyle = "Mono"
 
-func plotMeanVals(vals []stats.BenchRound, title string, fes int) *plot.Plot {
+func plotMeanVals(meanVals []float64, title string, fes int) *plot.Plot {
 	plotter.DefaultFont.Typeface = preferredFontStyle
 	plotter.DefaultLineStyle.Width = vg.Points(2.0)
 
@@ -43,9 +43,6 @@ func plotMeanVals(vals []stats.BenchRound, title string, fes int) *plot.Plot {
 	p.Title.TextStyle.Font.Weight = 2 // SemiBold
 	p.Title.Padding = 3 * vg.Millimeter
 
-	// get mean vals.
-	meanVals := stats.GetMeanVals(vals, fes)
-
 	// mark the end of the X axis with len(meanVals).
 	p.X.Max = float64(len(meanVals))
 	p.Y.Min = floats.Min(meanVals)
@@ -190,7 +187,7 @@ func plotAllDims(algoStats []stats.Stats, fPrefix, fExt string, ch chan report.P
 			picMean.FilePath = filenameMean
 
 			// get the *mean* plot.
-			pMean := plotMeanVals(dim.BenchResults, meanTitle, s.Generations)
+			pMean := plotMeanVals(dim.MeanVals, meanTitle, s.Generations)
 
 			elapsed := time.Since(start)
 			info := "saving img to file: " + filename + "(-Mean)" + fExt +

algo/randomSearch.go

@@ -99,6 +99,8 @@ func RandomSearchNG(maxFES, benchMinIters int, theD []int, benchFunc string, ch
 		)),
 	}
 
+	rsMeans := &stats.AlgoMean{Algo: "Random Search"}
+
 	// iterate over whatever was passed to us with theD - dimens slice.
 	for _, dimens := range localD {
 		randomSearchStatDimX := &stats.Stats{
@@ -109,6 +111,14 @@ func RandomSearchNG(maxFES, benchMinIters int, theD []int, benchFunc string, ch
 		}
 		funcStats := &stats.FuncStats{BenchName: benchFunc}
 		benchFuncParams := bench.FunctionParams[benchFunc]
+		dimXMean := &stats.Mean{
+			Bench:       benchFunc,
+			Dimens:      dimens,
+			Iterations:  minIters,
+			Generations: fes,
+			// not applicable to Random Search...
+			Neighbours: -1,
+		}
 
 		// set min/max bounds.
 		uniformDist.Min = benchFuncParams.Min()
@@ -156,6 +166,11 @@ func RandomSearchNG(maxFES, benchMinIters int, theD []int, benchFunc string, ch
 			}
 		}
 
+		// get mean vals.
+		dimXMean.MeanVals = stats.GetMeanVals(funcStats.BenchResults, fes)
+		// save to funcStats, too.
+		funcStats.MeanVals = dimXMean.MeanVals
+
 		// save cumulative results of 'minIters' runs.
 		randomSearchStatDimX.BenchFuncStats = append(
 			randomSearchStatDimX.BenchFuncStats,
@@ -164,7 +179,15 @@ func RandomSearchNG(maxFES, benchMinIters int, theD []int, benchFunc string, ch
 
 		// save stats for each dimension to a stats slice.
 		randomSearchStats = append(randomSearchStats, *randomSearchStatDimX)
+
+		// save to AlgoMeans
+		rsMeans.Means = append(rsMeans.Means, *dimXMean)
 	}
 
+	// export AlgoMeans.
+	mu.Lock()
+	meanStats.AlgoMeans = append(meanStats.AlgoMeans, *rsMeans)
+	mu.Unlock()
+
 	ch <- randomSearchStats
 }

algo/stochasticHillClimbing.go

@@ -186,6 +186,8 @@ func HillClimb(maxFES, benchMinIters int, theD []int, benchFunc string, ch chan
 	localD = theD
 	minIters = benchMinIters
 
+	shcMeans := &stats.AlgoMean{Algo: "Stochastic Hill Climbing"}
+
 	for _, dimens := range localD {
 		stochasticHCStatDimX := &stats.Stats{
 			Algo:       "Stochastic Hill Climbing",
@@ -198,6 +200,13 @@ func HillClimb(maxFES, benchMinIters int, theD []int, benchFunc string, ch chan
 		}
 		funcStats := &stats.FuncStats{BenchName: benchFunc}
 		benchFuncParams := bench.FunctionParams[benchFunc]
+		dimXMean := &stats.Mean{
+			Bench:       benchFunc,
+			Dimens:      dimens,
+			Iterations:  minIters,
+			Generations: fesPerIter,
+			Neighbours:  bench.Neighbourhood,
+		}
 		uniDist := distuv.Uniform{
 			Min: benchFuncParams.Min(),
 			Max: benchFuncParams.Max(),
@@ -276,13 +285,26 @@ func HillClimb(maxFES, benchMinIters int, theD []int, benchFunc string, ch chan
 			benchFunc + "\" for " + fmt.Sprint(dimens) + "D took " + fmt.Sprint(elapsed),
 		)
 
+		// get mean vals.
+		dimXMean.MeanVals = stats.GetMeanVals(funcStats.BenchResults, fesPerIter)
+		// save to funcStats, too.
+		funcStats.MeanVals = dimXMean.MeanVals
+
 		stochasticHCStatDimX.BenchFuncStats = append(
 			stochasticHCStatDimX.BenchFuncStats,
 			*funcStats,
 		)
 
 		stochasticHCStats = append(stochasticHCStats, *stochasticHCStatDimX)
+
+		// save to AlgoMeans.
+		shcMeans.Means = append(shcMeans.Means, *dimXMean)
 	}
 
+	// export AlgoMeans.
+	mu.Lock()
+	meanStats.AlgoMeans = append(meanStats.AlgoMeans, *shcMeans)
+	mu.Unlock()
+
 	ch <- stochasticHCStats
 }

stats/stats.go

@@ -21,6 +21,34 @@ type BenchRound struct {
 type FuncStats struct {
 	BenchName    string
 	BenchResults []BenchRound
+	// MeanVals holds the mean value for each step of the iteration.
+	MeanVals []float64
+}
+
+// Mean structure holds mean vals and metadata of a specified
+// bench-algo-settings combination.
+type Mean struct {
+	Bench  string
+	Dimens int
+
+	Iterations  int
+	Generations int
+	// Neighbours param is used in Hill Climbing type algos, set to -1 if not
+	// applicable.
+	Neighbours int
+	// MeanVals holds the mean value for each step of the iteration.
+	MeanVals []float64
+}
+
+// AlgoMean holds Mean structs of different benchmarks but the same algorithm.
+type AlgoMean struct {
+	Algo  string
+	Means []Mean
+}
+
+// MeanStats aggregates AlgoMean structs of different algos.
+type MeanStats struct {
+	AlgoMeans []AlgoMean
 }
 
 type Stats struct {