-
Notifications
You must be signed in to change notification settings - Fork 3
/
Copy path09_get_VAE_distribution.R
184 lines (161 loc) · 7.26 KB
/
09_get_VAE_distribution.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
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
175
176
177
178
179
180
181
182
183
184
# Get distribution of clinical parameter p-values at finer granularity
# Load encodings ----------------------
vae_reps_path <- "results/replications"
pca_path <- 'results/encodings/'
vae_reps_files <-
list.files(vae_reps_path,
recursive = TRUE,
full.names = TRUE)
d_dims <- str_extract(vae_reps_files, "\\d{2}[_]") %>%
unique() %>%
strsplit("_") %>%
unlist()
models<- c("PCA", "RBF KPCA", "Sigmoid KPCA", "Polynomial KPCA", "Cosine KPCA")
# Load PCA top performing dimensions --------------------------------------------------
# NOTE: This code will only run if you have created encodings
# for the PCA and KPCA models at d = [10,13,15,16,17,18,20]
top_performers_pca<-
d_dims %>% lapply(function(dim){
qm_encoding <-
bind_rows(
glue("{pca_path}//QMDiab_PCA_encoding_d{dim}.csv") %>%
read_csv() %>%
process_qmdiab() %>%
dplyr::mutate(model = "PCA"),
glue("{pca_path}//QMDiab_KPCA_cosine_encoding_d{dim}.csv") %>%
read_csv() %>%
process_qmdiab() %>%
dplyr::mutate(model = "Cosine KPCA"),
glue("{pca_path}//QMDiab_KPCA_sigmoid_encoding_d{dim}.csv") %>%
read_csv() %>%
process_qmdiab() %>%
dplyr::mutate(model = "Sigmoid KPCA"),
glue("{pca_path}//QMDiab_KPCA_rbf_encoding_d{dim}.csv") %>%
read_csv() %>%
process_qmdiab() %>%
dplyr::mutate(model = "RBF KPCA"),
glue("{pca_path}//QMDiab_KPCA_poly_encoding_d{dim}.csv") %>%
read_csv() %>%
process_qmdiab() %>%
dplyr::mutate(model = "Polynomial KPCA")
)
qm_scores <- qm_encoding %>% get_latent_scores()
qm_mins <- models %>% lapply(function(mod){log10(min(qm_scores %>%
filter(model==mod) %>% .$pval_nominal))})
qm_df <- tibble(model = models,
dim = rep(dim, times = length(models)),
disease = rep("qm", times= length(models)),
mins = qm_mins %>% unlist() )#%>% log10())
aml_encoding <-
bind_rows(
glue("{pca_path}//AML_PCA_encoding_d{dim}.csv") %>%
read_csv() %>%
process_aml("^[0-9]+") %>%
dplyr::mutate(model = "PCA"),
glue("{pca_path}//AML_KPCA_cosine_encoding_d{dim}.csv") %>%
read_csv() %>%
process_aml("^[0-9]+") %>%
dplyr::mutate(model = "Cosine KPCA"),
glue("{pca_path}//AML_KPCA_sigmoid_encoding_d{dim}.csv") %>%
read_csv() %>%
process_aml("^[0-9]+") %>%
dplyr::mutate(model = "Sigmoid KPCA"),
glue("{pca_path}//AML_KPCA_RBF_encoding_d{dim}.csv") %>%
read_csv() %>%
process_aml("^[0-9]+") %>%
dplyr::mutate(model = "RBF KPCA"),
glue("{pca_path}//AML_KPCA_poly_encoding_d{dim}.csv") %>%
read_csv() %>%
process_aml("^[0-9]+") %>%
dplyr::mutate(model = "Polynomial KPCA")
)
aml_scores <- aml_encoding %>% get_latent_scores()
aml_mins <- models %>% lapply(function(mod){log10(min(aml_scores %>%
filter(model==mod) %>% .$pval_nominal))})
aml_df <- tibble(model = models,
dim = rep(dim, times = length(models)),
disease = rep("aml", times= length(models)),
mins = aml_mins %>% unlist() )#%>% log10())
schizo_encoding <-
bind_rows(
glue("{pca_path}//Schizo_PCA_encoding_d{dim}.csv") %>%
read_csv() %>%
process_schizo("^[0-9]+") %>%
dplyr::mutate(model = "PCA"),
glue("{pca_path}//Schizo_KPCA_cosine_encoding_d{dim}.csv") %>%
read_csv() %>%
process_schizo("^[0-9]+") %>%
dplyr::mutate(model = "Cosine KPCA"),
glue("{pca_path}//Schizo_KPCA_sigmoid_encoding_d{dim}.csv") %>%
read_csv() %>%
process_schizo("^[0-9]+") %>%
dplyr::mutate(model = "Sigmoid KPCA"),
glue("{pca_path}//Schizo_KPCA_rbf_encoding_d{dim}.csv") %>%
read_csv() %>%
process_schizo("^[0-9]+") %>%
dplyr::mutate(model = "RBF KPCA"),
glue("{pca_path}//Schizo_KPCA_poly_encoding_d{dim}.csv") %>%
read_csv() %>%
process_schizo("^[0-9]+") %>%
dplyr::mutate(model = "Polynomial KPCA")
)
schizo_scores <- schizo_encoding %>% get_latent_scores()
schizo_mins <- models %>% lapply(function(mod){log10(min(schizo_scores %>%
filter(model==mod) %>% .$pval_nominal))})
schizo_df <-tibble(model = models,
dim = rep(dim, times = length(models)),
disease = rep("schizo", times= length(models)),
mins = schizo_mins %>% unlist() )#%>% log10())
bind_rows(qm_df, aml_df, schizo_df)
}) %>% bind_rows()
# Load VAE top performing dimensions --------------------------------------------------
# NOTE: This code will only run if you have trained a VAE model
# at d = [10,13,15,16,17,18,20] over multiple replications
reps <- str_extract(vae_reps_files, "[0-9]{1,3}.csv") %>% unique()
top_performers_vae<-
d_dims %>% lapply(function(dim){
reps %>% lapply(function(rep){
qm_encoding <-
glue("{vae_reps_path}/QMDiab_VAE_encoding_{dim}_{rep}") %>%
read_csv() %>%
process_qmdiab() %>%
dplyr::mutate(model = "VAE")
qm_scores <- qm_encoding %>% get_latent_scores()
qm_min <- min(qm_scores %>% .$pval_nominal)
aml_encoding <-
glue("{vae_reps_path}/AML_VAE_encoding_{dim}_{rep}") %>%
read_csv() %>%
process_aml("^[0-9]+") %>%
dplyr::mutate(model = "VAE")
aml_scores <- aml_encoding %>% get_latent_scores()
aml_min <- min(aml_scores %>% .$pval_nominal)
schizo_encoding <-
glue("{vae_reps_path}/Schizo_VAE_encoding_{dim}_{rep}") %>%
read_csv() %>%
process_schizo("^[0-9]+") %>%
dplyr::mutate(model = "VAE")
schizo_scores <- schizo_encoding %>% get_latent_scores()
schizo_min <- min(schizo_scores %>% .$pval_nominal)
rep_df <- tibble(rep = rep(rep,3),
model = rep("VAE", 3),
dim = rep(dim, 3),
disease = c("qm", "aml", "schizo"),
vae_mins = c(qm_min, aml_min, schizo_min))
}) %>% bind_rows()
}) %>% bind_rows()
# Calculate error bars and plot --------------------------------------------------
df<- top_performers_vae %>% group_by(dim,disease) %>%
dplyr::summarise(mean = mean(log10(vae_mins)),
sd = sd(log10(vae_mins)),
min = mean-sd,
max = mean+sd)
ggplot(df,aes(x = dim, y =mean))+
geom_line(aes(group=1))+geom_errorbar(aes(ymin=min, ymax = max),width = 2, color = "black")+
geom_line(data=top_performers_pca,aes(x=dim, y=mins, group = model,color=model))+facet_wrap(~disease,scale="free")+
theme_minimal() +
theme(axis.title.x = element_blank(),
axis.title.y = element_blank(),
axis.ticks.x = element_blank(),
panel.spacing = unit(2, "lines"),
strip.text.y = element_blank()) +
scale_color_manual(values=brewer.pal(6, name = "Dark2"))