-
Notifications
You must be signed in to change notification settings - Fork 27
/
vis_model_contours.sh
executable file
·49 lines (43 loc) · 2.7 KB
/
vis_model_contours.sh
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
#!/bin/sh
#MLE
python train_syn.py --cost_type=MLE \
--dataset_filename=syndata_1.pkl --save_filename=MLE_1.pkl
python train_syn.py --cost_type=MLE \
--dataset_filename=syndata_2.pkl --save_filename=MLE_2.pkl
python visualize_contour.py --load_filename=MLE_1.pkl --save_filename=MLE_1.pdf --dataset_i=1
python visualize_contour.py --load_filename=MLE_2.pkl --save_filename=MLE_2.pdf --dataset_i=2
#L2
python train_syn.py --cost_type=L2 --lamb=0.0001 \
--dataset_filename=syndata_1.pkl --save_filename=L2_1.pkl
python train_syn.py --cost_type=L2 --lamb=0.0001 \
--dataset_filename=syndata_2.pkl --save_filename=L2_2.pkl
python visualize_contour.py --load_filename=L2_1.pkl --save_filename=L2_1.pdf --dataset_i=1
python visualize_contour.py --load_filename=L2_2.pkl --save_filename=L2_2.pdf --dataset_i=2
#dropout
python train_syn.py --cost_type=dropout --dropout_rate=0.4 \
--dataset_filename=syndata_1.pkl --save_filename=dropout_1.pkl
python train_syn.py --cost_type=dropout --dropout_rate=0.75 \
--dataset_filename=syndata_2.pkl --save_filename=dropout_2.pkl
python visualize_contour.py --load_filename=dropout_1.pkl --save_filename=dropout_1.pdf --dataset_i=1
python visualize_contour.py --load_filename=dropout_2.pkl --save_filename=dropout_2.pdf --dataset_i=2
#Random Perturbation training
python train_syn.py --cost_type=VAT --num_power_iter=0 --epsilon=1.8 \
--dataset_filename=syndata_1.pkl --save_filename=RP_1.pkl
python train_syn.py --cost_type=VAT --num_power_iter=0 --epsilon=2.8 \
--dataset_filename=syndata_2.pkl --save_filename=RP_2.pkl
python visualize_contour.py --load_filename=RP_1.pkl --save_filename=RP_1.pdf --dataset_i=1
python visualize_contour.py --load_filename=RP_2.pkl --save_filename=RP_2.pdf --dataset_i=2
#Adversarial training
python train_syn.py --cost_type=AT --epsilon=0.2 \
--dataset_filename=syndata_1.pkl --save_filename=AT_1.pkl
python train_syn.py --cost_type=AT --epsilon=0.65 \
--dataset_filename=syndata_2.pkl --save_filename=AT_2.pkl
python visualize_contour.py --load_filename=AT_1.pkl --save_filename=AT_1.pdf --dataset_i=1
python visualize_contour.py --load_filename=AT_2.pkl --save_filename=AT_2.pdf --dataset_i=2
#Virtual Adversarial training
python train_syn.py --cost_type=VAT_finite_diff --num_power_iter=1 --epsilon=0.6 \
--dataset_filename=syndata_1.pkl --save_filename=VAT_1.pkl
python train_syn.py --cost_type=VAT_finite_diff --num_power_iter=1 --epsilon=0.5 \
--dataset_filename=syndata_2.pkl --save_filename=VAT_2.pkl
python visualize_contour.py --load_filename=VAT_1.pkl --save_filename=VAT_1.pdf --dataset_i=1
python visualize_contour.py --load_filename=VAT_2.pkl --save_filename=VAT_2.pdf --dataset_i=2