-
-
Notifications
You must be signed in to change notification settings - Fork 9
/
knn.go
174 lines (147 loc) · 4.07 KB
/
knn.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
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
package knn
import (
"encoding/gob"
"os"
"path"
"github.com/jaimeteb/chatto/internal/clf/dataset"
"github.com/jaimeteb/chatto/internal/clf/pipeline"
"github.com/jaimeteb/chatto/internal/clf/wordvectors"
"github.com/mitchellh/mapstructure"
log "github.com/sirupsen/logrus"
)
const (
classifierFile = "clf.gob"
modelFile = "knn.gob"
)
// Classifier is a K-Nearest Neighbors classifier
type Classifier struct {
KNN *KNN
VectorMap *wordvectors.VectorMap
K int
}
// Parameters represents the model hyperparameters
type Parameters struct {
K int `mapstructure:"k"`
}
func (c *Classifier) SaveToFile(name string) error {
file, err := os.OpenFile(name, os.O_WRONLY|os.O_CREATE, 0644)
if err != nil {
return err
}
defer file.Close()
enc := gob.NewEncoder(file)
return enc.Encode(&Classifier{nil, nil, c.K})
}
func NewClassifierFromFile(name string) (*Classifier, error) {
file, err := os.Open(name)
if err != nil {
return nil, err
}
defer file.Close()
dec := gob.NewDecoder(file)
classifier := new(Classifier)
err = dec.Decode(classifier)
return classifier, err
}
// NewClassifier creates a KNN classifier with truncate and file data
func NewClassifier(wordVecConfig wordvectors.Config, params map[string]interface{}) *Classifier {
decParams := &Parameters{
K: 1,
}
dec, err := mapstructure.NewDecoder(&mapstructure.DecoderConfig{
Metadata: nil,
Result: decParams,
})
if err != nil || dec.Decode(params) != nil {
log.Errorf("Could not parse parameters: %v", err)
}
// Generate VectorMap
emb, err := wordvectors.NewVectorMap(&wordVecConfig)
if err != nil {
log.Fatal(err)
}
return &Classifier{
VectorMap: emb,
K: decParams.K,
}
}
// Learn takes the training texts and trains the K-Nearest Neighbors classifier
func (c *Classifier) Learn(texts dataset.DataSet, pipe *pipeline.Config) float32 {
trainX := make([][]string, 0)
trainY := make([]string, 0)
// Run Pipeline
for _, training := range texts {
for _, trainingText := range training.Texts {
trainX = append(trainX, pipeline.Pipeline(trainingText, pipe))
trainY = append(trainY, training.Command)
}
}
// Get embeddings from dataset
embeddingsX := make([][]float64, len(trainX))
for i, x := range trainX {
embeddingsX[i] = c.VectorMap.AverageVectors(c.VectorMap.Vectors(x))
}
// Initialize KNN
knn := &KNN{
K: c.K,
Data: embeddingsX,
Labels: trainY,
}
c.KNN = knn
// Compute train accuracy
preds, _ := c.KNN.PredictMany(embeddingsX)
correct := 0
for i, pred := range preds {
if pred == trainY[i] {
correct++
}
}
return float32(correct) / float32(len(preds))
}
// Predict predict a class for a given text
func (c *Classifier) Predict(text string, pipe *pipeline.Config) (predictedClass string, proba float32) {
x := pipeline.Pipeline(text, pipe)
embeddingsX := c.VectorMap.AverageVectors(c.VectorMap.Vectors(x))
pred, prob := c.KNN.PredictOne(embeddingsX)
log.Debugf("CLF | Text '%s' classified as command '%s' with a probability of %.2f", text, pred, prob)
if prob < pipe.Threshold {
return "", -1.0
}
return pred, float32(prob)
}
// Save persists the model to a file
func (c *Classifier) Save(directory string) error {
// save Classifier
if err := c.SaveToFile(path.Join(directory, classifierFile)); err != nil {
return err
}
// save Classifier.VectorMap
if err := c.VectorMap.SaveToFile(path.Join(directory, wordvectors.WordVectorsFile)); err != nil {
return err
}
// save Classifier.KNN
if err := c.KNN.SaveToFile(path.Join(directory, modelFile)); err != nil {
return err
}
return nil
}
func Load(directory string) (classifier *Classifier, err error) {
// load Classifier
classifier, err = NewClassifierFromFile(path.Join(directory, classifierFile))
if err != nil {
return
}
// load Classifier.VectorMap
vectorMap, err := wordvectors.NewVectorMapFromFile(path.Join(directory, wordvectors.WordVectorsFile))
if err != nil {
return
}
classifier.VectorMap = vectorMap
// load Classifier.KNN
knn, err := NewKNNClassifierFromFile(path.Join(directory, modelFile))
if err != nil {
return
}
classifier.KNN = knn
return
}