func (a *AnnTrainer) Predict(sample *core.Sample) {
x := core.NewVector()
y := core.NewVector()
//z := core.NewVector()
for i := ZEROI64; i < a.Hidden[0]; i++ {
sum := ZEROF64
for _, f := range sample.Features {
sum += f.Value * a.Model[0].Data[i].GetValue(f.Id)
}
x.Data[i] = sigMoid(sum)
}
x.Data[a.Hidden[0]] = 1
for i := ZEROI64; i < a.Hidden[1]; i++ {
sum := ZEROF64
for j := ZEROI64; j <= a.Hidden[0]; j++ {
sum += x.GetValue(j) * a.Model[1].GetValue(j, i)
}
y.Data[i] = sigMoid(sum)
}
y.Data[a.Hidden[1]] = 1
sum := ZEROF64
for j := ZEROI64; j <= a.Hidden[1]; j++ {
sum += y.GetValue(j) * a.Model[2].GetValue(j, 0)
}
sample.Prediction = sum
}
func (a *AnnTrainer) LoadModel(path string) {
f, err := os.Open(path)
if err != nil {
fmt.Println(err)
return
}
byt, err := ioutil.ReadAll(f)
if err != nil {
fmt.Println(err)
return
}
lines := strings.Split(string(byt), "\n")
layerNum, err := strconv.Atoi(lines[0])
if err != nil {
fmt.Println(err)
return
}
a.Model = make([]*core.Matrix, layerNum)
lines = lines[1:]
var index int
for index < layerNum {
row, err := strconv.ParseInt(lines[0], 10, 64)
if err != nil {
fmt.Println(err)
return
}
col, err := strconv.ParseInt(lines[1], 10, 64)
if err != nil {
fmt.Println(err)
return
}
fmt.Printf("has %d * %d matrix\n", row, col)
lines = lines[2:]
a.Model[index] = core.NewMatrix()
for i := int64(0); i < row; i++ {
a.Model[index].Data[i] = core.NewVector()
for j := int64(0); j < col; j++ {
a.Model[index].SetValue(j, i, parseFloat64(lines[j]))
}
lines = lines[col:]
}
index++
}
}