// Reads a .dat file where the entries are matricies of numbers
// separated by empty new lines and individually aligned with
// whitespace between row entries and newlines between rows
func Read(file []byte) []matrix.Matrix {
data := make([]matrix.Matrix, 0)
for _, mat := range bytes.Split(file, []byte("\n\n")) {
temp := make(matrix.Matrix, 0)
didParse := true
Element:
for _, row := range bytes.Split(mat, []byte("\n")) {
floatsAsStrings := strings.Fields(string(row))
elms := make([]matrix.Element, 0)
for _, s := range floatsAsStrings {
f, err := strconv.ParseFloat(s, 64)
if err != nil {
didParse = false
break Element
}
elms = append(elms, matrix.Element(f))
}
temp = append(temp, elms)
}
if didParse && len(temp[0]) > 0 {
data = append(data, temp)
}
}
return data
}
func (g *generator) Distance() (matrix.Matrix, error) {
m := matrix.Matrix(make([][]matrix.Element, g.n))
for i := 0; i < g.n; i++ {
m[i] = make([]matrix.Element, g.n)
for j := 0; j < g.n; j++ {
m[i][j] = matrix.Element(rand.Float64())
}
}
return m, nil
}
func (g *generator) Flow(spread float64) (matrix.Matrix, error) {
if spread < 0 || spread >= 1 {
return nil, fmt.Errorf("Error: spread must be between 0 and 1.")
}
// Create Zipf generator
scale := 1000
r := rand.New(rand.NewSource(0))
zipf := rand.NewZipf(r, 1.01, float64(g.n), uint64(scale))
// Populate frequencies of unigrams
k := make([]float64, g.n)
for i := 0; i < g.n; i++ {
k[i] = float64(zipf.Uint64())
}
// Populate ideal bigram matrix
m := matrix.Matrix(make([][]matrix.Element, g.n))
for i := 0; i < g.n; i++ {
m[i] = make([]matrix.Element, g.n)
for j := 0; j < g.n; j++ {
e := (rand.Float64() - 0.5) * spread
m[i][j] = matrix.Element((k[i] * k[j]) * (1 + e))
}
}
// Scale back to 100,000 total freq
totalF := m.Sum()
s := g.fscale / totalF
for i := 0; i < g.n; i++ {
for j := 0; j < g.n; j++ {
m[i][j] = matrix.Element(math.Floor(float64(m[i][j]) * s))
}
}
return m, nil
}
