microcanonical entropy

cryptipass.go

@@ -246,7 +246,7 @@ func (g *Generator) GenNextToken(seed string) (string, float64) {
 			N := g.Rng.IntN(tr.total)
 			for i, v := range tr.counts {
 				if N < v {
-					return tr.tokens[i], tr.entropy
+					return tr.tokens[i], tr.entropies[i]
 				}
 			}
 			panic("unexpected")
@@ -280,7 +280,7 @@ func (g *Generator) GenWordLength() (int, float64) {
 		N := g.Rng.IntN(tr.total)
 		for i, v := range tr.counts {
 			if N < v {
-				return int(tr.tokens[i][0]), tr.entropy
+				return int(tr.tokens[i][0]), tr.entropies[i]
 			}
 		}
 	}

internal.go

@@ -45,10 +45,10 @@ func (g *Generator) assert_ready() {
 // The distribution struct is used internally to facilitate the generation
 // of passwords by modeling the relationships between characters in the generated output.
 type distribution struct {
-	tokens  []string
-	counts  []int
-	total   int
-	entropy float64
+	tokens    []string
+	counts    []int
+	entropies []float64
+	total     int
 }
 
 // distill processes a list of words and builds a probabilistic transition matrix
@@ -105,21 +105,19 @@ func distill(tokens []string, depth int) map[string]distribution {
 		for _, freq := range rfreq {
 			C += freq
 		}
-		H := 0.0
 		tr := distribution{}
 		tr.counts = make([]int, 0)
 
 		tr.tokens = make([]string, 0)
 		cum := 0
 		for ru, freq := range rfreq {
 			p := float64(freq) / float64(C)
-			H -= math.Log2(p) * p
 			cum += freq
 			tr.counts = append(tr.counts, cum)
 			tr.tokens = append(tr.tokens, string(ru))
+			tr.entropies = append(tr.entropies, math.Log2(1.0/p))
 		}
 		tr.total = C
-		tr.entropy = H
 		dist_trans_matrix[k] = tr
 	}
 	return dist_trans_matrix