-
Notifications
You must be signed in to change notification settings - Fork 2k
/
runit_deeplearning_weights_and_biases.R
80 lines (59 loc) · 3.09 KB
/
runit_deeplearning_weights_and_biases.R
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
setwd(normalizePath(dirname(R.utils::commandArgs(asValues=TRUE)$"f")))
source("../../../scripts/h2o-r-test-setup.R")
check.deeplearning_imbalanced <- function() {
Log.info("Test checks if Deep Learning weights and biases are accessible from R")
census <- h2o.uploadFile(locate("smalldata/chicago/chicagoCensus.csv"))
census[,1] <- as.factor(census[,1])
dlmodel<-h2o.deeplearning(x=c(1:3),y=4,hidden=c(17,191),epochs=1,
training_frame=census,balance_classes=F,
reproducible=T, seed=1234, export_weights_and_biases=T)
#print(dlmodel)
weights1 <- h2o.weights(dlmodel,matrix_id=1)
print(head(weights1))
weights2 <- h2o.weights(dlmodel,matrix_id=2)
weights3 <- h2o.weights(dlmodel,matrix_id=3)
biases1 <- h2o.biases(dlmodel,vector_id=1)
biases2 <- h2o.biases(dlmodel,vector_id=2)
biases3 <- h2o.biases(dlmodel,vector_id=3)
checkTrue(ncol(weights1) == 79, "wrong dimensionality!")
checkTrue(nrow(weights1) == 17, "wrong dimensionality!")
checkTrue(ncol(weights2) == 17, "wrong dimensionality!")
checkTrue(nrow(weights2) == 191, "wrong dimensionality!")
checkTrue(ncol(weights3) == 191, "wrong dimensionality!")
checkTrue(nrow(weights3) == 1, "wrong dimensionality!")
checkTrue(ncol(biases1) == 1, "wrong dimensionality!")
checkTrue(nrow(biases1) == 17, "wrong dimensionality!")
checkTrue(ncol(biases2) == 1, "wrong dimensionality!")
checkTrue(nrow(biases2) == 191, "wrong dimensionality!")
checkTrue(ncol(biases3) == 1, "wrong dimensionality!")
checkTrue(nrow(biases3) == 1, "wrong dimensionality!")
df <- as.h2o(iris)
dl1 <- h2o.deeplearning(1:4,5,df,hidden=c(10,10),export_weights_and_biases = TRUE, seed=1234, reproducible=T)
p1 <- h2o.predict(dl1, df)
ll1 <- h2o.logloss(h2o.performance(dl1,df))
print(ll1)
## get weights and biases
w1 <- h2o.weights(dl1,1)
w2 <- h2o.weights(dl1,2)
w3 <- h2o.weights(dl1,3)
b1 <- h2o.biases(dl1,1)
b2 <- h2o.biases(dl1,2)
b3 <- h2o.biases(dl1,3)
## make a model from given weights/biases
dl2 <- h2o.deeplearning(1:4,5,df,hidden=c(10,10),initial_weights=c(w1,w2,w3),initial_biases=c(b1,b2,b3), epochs=0)
p2 <- h2o.predict(dl2, df)
ll2 <- h2o.logloss(h2o.performance(dl2,df))
print(ll2)
#h2o.download_pojo(dl2) ## fully functional pojo
## check consistency
checkTrue(max(abs(p1[,2:4]-p2[,2:4])) < 1e-6)
checkTrue(abs(ll2 - ll1) < 1e-6)
## make another model with partially set weights/biases
dl3 <- h2o.deeplearning(1:4,5,training_frame=df,hidden=c(10,10),initial_weights=list(w1,NULL,w3),initial_biases=list(b1,b2,NULL), epochs=10, seed=1234, reproducible=T)
ll3 <- h2o.logloss(h2o.performance(dl3,df))
checkTrue(ll3 < ll1)
## make another model with partially set user-modified weights/biases
dl4 <- h2o.deeplearning(1:4,5,training_frame=df,hidden=c(10,10),initial_weights=list(w1*1.1,w2*0.9,sqrt(w3)),initial_biases=list(b1,b2,NULL), epochs=10, seed=1234, reproducible=T)
ll4 <- h2o.logloss(h2o.performance(dl4,df))
}
doTest("Deep Learning Weights/Biases Test", check.deeplearning_imbalanced)