-
Notifications
You must be signed in to change notification settings - Fork 5
/
config_real_datasets.py
executable file
·67 lines (51 loc) · 2.17 KB
/
config_real_datasets.py
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
#Some important constants
RANDOM_SEED = 35813
#Some important paths
AD_LH_DATASET_PATH = "./dataset/data_ad_lmci_L/AD LH"
AD_RH_DATASET_PATH = "./dataset/data_ad_lmci_R/AD RH"
LMCI_LH_DATASET_PATH = "./dataset/data_ad_lmci_L/LMCI LH"
LMCI_RH_DATASET_PATH = "./dataset/data_ad_lmci_R/LMCI RH"
NC_LH_DATASET_PATH = "./dataset/data_nc_asd_L/NC LH"
NC_RH_DATASET_PATH = "./dataset/data_nc_asd_R/NC RH"
ASD_LH_DATASET_PATH = "./dataset/data_nc_asd_L/ASD LH"
ASD_RH_DATASET_PATH = "./dataset/data_nc_asd_R/ASD RH"
MODEL_WEIGHT_BACKUP_PATH = "./output/model_weights"
DEEP_CBT_SAVE_PATH = "./output/deep_cbts"
TEMP_FOLDER = "./temp"
#Model Configuration
N_ROIs = 35
N_RANDOM_SAMPLES = 10
#AD_LMCI KL LOSS
KL_AD_LMCI_Nattr = 4
KL_AD_LMCI_CONV1 = 36
KL_AD_LMCI_CONV2 = 24
KL_AD_LMCI_CONV3 = 5
AD_LMCI_lambda_kl = 25
KL_AD_LMCI_PARAMS = {
"learning_rate" : 0.0006,
"n_attr": KL_AD_LMCI_Nattr,
"lambda_kl": AD_LMCI_lambda_kl,
"Linear1" : {"in": KL_AD_LMCI_Nattr, "out": KL_AD_LMCI_CONV1},
"conv1": {"in" : 1, "out": KL_AD_LMCI_CONV1},
"Linear2" : {"in": KL_AD_LMCI_Nattr, "out": KL_AD_LMCI_CONV1 * KL_AD_LMCI_CONV2},
"conv2": {"in" : KL_AD_LMCI_CONV1, "out": KL_AD_LMCI_CONV2},
"Linear3" : {"in": KL_AD_LMCI_Nattr, "out": KL_AD_LMCI_CONV2 * KL_AD_LMCI_CONV3},
"conv3": {"in" : KL_AD_LMCI_CONV2, "out": KL_AD_LMCI_CONV3},
}
#NC_ASD KL LOSS
KL_NC_ASD_Nattr = 6
KL_NC_ASD_CONV1 = 36
KL_NC_ASD_CONV2 = 24
KL_NC_ASD_CONV3 = 8
NC_ASD_lambda_kl = 10
KL_NC_ASD_PARAMS = {
"learning_rate" : 0.0006,
"n_attr": KL_NC_ASD_Nattr,
"lambda_kl": NC_ASD_lambda_kl,
"Linear1" : {"in": KL_NC_ASD_Nattr, "out": KL_NC_ASD_CONV1},
"conv1": {"in" : 1, "out": KL_NC_ASD_CONV1},
"Linear2" : {"in": KL_NC_ASD_Nattr, "out": KL_NC_ASD_CONV1 * KL_NC_ASD_CONV2},
"conv2": {"in" : KL_NC_ASD_CONV1, "out": KL_NC_ASD_CONV2},
"Linear3" : {"in": KL_NC_ASD_Nattr, "out": KL_NC_ASD_CONV2 * KL_NC_ASD_CONV3},
"conv3": {"in" : KL_NC_ASD_CONV2, "out": KL_NC_ASD_CONV3},
}