forked from blei-lab/deep-exponential-families
-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathdef_wikipedia_50_25_10.ini
106 lines (94 loc) · 2.34 KB
/
def_wikipedia_50_25_10.ini
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
; A three layer Gamma-Poisson DEF
; for the wikipedia dataset
; Strings in square brackets (i.e., []) indicate sections
; the max number of examples for serializing z
serialization_max_examples=10000
[global]
min_gamma_shape=0.005
min_gamma_scale=1e-5
min_gamma_sample=1e-300
; Data Sections (train/valid/test).
; Each section describes the type of data to be modelled and its physical location
[train]
data_type=sparse-text
data_file=${WIKIPEDIA_DEF}/train.cpp.dat
;Deserialize an existing model (useful to load a previously trained
;model). Can be be used in any data section.
;deserialize=
[valid]
data_type=masked-text
data_file=${WIKIPEDIA_DEF}/validation.cpp.dat
; enforce that validation/test model must have factorized layer on the first layer
layer0.q_type=factorized
[test]
data_type=masked-text
data_file=${WIKIPEDIA_DEF}/test.cpp.dat
; enforce that validation/test model must have factorized layer on the first layer
layer0_q_type=factorized
; parameter for all weights layer
[DEF_weights]
; specifications of the latent weights variables
type=gamma
; prior gamma parameters
shape=0.1
scale=3.3
q_lf=softmax
; wshape = exp(gaussian) * wshape_init
wshape_init=0.01
wscale_init=0.01
[obs_weights]
; specification of the latent weights variables (lowest layer)
type=gamma
; prior gamma parameters
shape=0.1
scale=3.3
q_lf=softmax
; wshape = exp(gaussian) * wshape_init
wshape_init=0.01
wscale_init=0.01
[y_layer]
; parameter of the prior distribution over the observations
type=poisson
; no intercept needed for gamma-poisson model
poisson_rate_intercept=0.0
; link function
lf=id
[DEF_layer1]
; specification of the Z variables at the first/bottom layer
size=50
type=gamma
; prior gamma parameters
shape=0.3
scale=3.3
q_lf=softmax
q_type=factorized
; how to initialize the variational approximation
wshape_init=0.01
wscale_init=0.01
lf=id
min_gamma_scale=1e-6
[DEF_layer2]
; specification of the Z variables at the second/middle layer
size=25
; prior gamma parameters
type=gamma
shape=0.3
scale=3.3
q_lf=softmax
q_type=factorized
; how to initialize the variational approximation
wshape_init=0.01
wscale_init=0.01
lf=id
min_gamma_scale=1e-6
[DEF_prior_layer]
; specification of the Z variables at the third/top layer
size=10
type=gamma
shape=0.3
scale=3.3
q_lf=softmax
q_type=factorized
; how to initialize the variational approximation
wshape_init=0.01
wscale_init=0.01