forked from ryanbressler/CloudForest
-
Notifications
You must be signed in to change notification settings - Fork 3
/
l1target.go
62 lines (51 loc) · 1.66 KB
/
l1target.go
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
package CloudForest
import (
"math"
)
/*
L1Target wraps a numerical feature as a target for us in l1 norm regression.
*/
type L1Target struct {
NumFeature
}
/*
L1Target.SplitImpurity is an L1 version of SplitImpurity.
*/
func (target *L1Target) SplitImpurity(l *[]int, r *[]int, m *[]int, allocs *BestSplitAllocs) (impurityDecrease float64) {
nl := float64(len(*l))
nr := float64(len(*r))
nm := 0.0
impurityDecrease = nl * target.Impurity(l, nil)
impurityDecrease += nr * target.Impurity(r, nil)
if m != nil && len(*m) > 0 {
nm = float64(len(*m))
impurityDecrease += nm * target.Impurity(m, nil)
}
impurityDecrease /= nl + nr + nm
return
}
//UpdateSImpFromAllocs willl be called when splits are being built by moving cases from r to l as in learning from numerical variables.
//Here it just wraps SplitImpurity but it can be implemented to provide further optimization.
func (target *L1Target) UpdateSImpFromAllocs(l *[]int, r *[]int, m *[]int, allocs *BestSplitAllocs, movedRtoL *[]int) (impurityDecrease float64) {
return target.SplitImpurity(l, r, m, allocs)
}
//L1Target.Impurity is an L1 version of impurity returning L1 instead of squared error.
func (target *L1Target) Impurity(cases *[]int, counter *[]int) (e float64) {
m := target.Mean(cases)
e = target.Error(cases, m)
return
}
//L1Target.MeanL1Error returns the Mean L1 norm error of the cases specified vs the predicted
//value. Only non missing cases are considered.
func (target *L1Target) Error(cases *[]int, predicted float64) (e float64) {
e = 0.0
n := 0
for _, i := range *cases {
if !target.IsMissing(i) {
e += math.Abs(predicted - target.Get(i))
n += 1
}
}
e = e / float64(n)
return
}