lls-ddfa/experiment_config.yml at main · manleyroberts/lls-ddfa · GitHub

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
# Indicate the preferred GPU for experiments, or -1 for no GPU.
gpu: 4

datasets:
  cifar10:
    # Variants of CIFAR do not require you to download the zip yourself, just indicate a new root and they will automatically download
    root_path: '~/cifar10/'
  cifar20:
    root_path: '~/cifar100/'
  imagenet:
    # ImageNet50 must be downloaded on your own. Please indicate the root directory of the dataset.
    root_path: '~/imagenet/'
  fg2:
    # FieldGuide2 must be installed on your own. Please indicate the root directory of the dataset.
    root_path: '~/fieldguide2/'
  fg28:
    # FieldGuide28 must be installed on your own. Please indicate the root directory of the dataset.
    root_path: '~/fieldguide28/'

# List of all experiments to conduct. If an experiment is commented out, it will not be conducted.
# Note: this is not an exhaustive list of the experiments conducted in our paper.
# Instead, this is designed to be a template for conducting a user's own experiments by modifying the following list.
experiments:
- dataset_settings:
    dataset: 'cifar10'
    dataset_split_seed: 7636
  class_prior_generation:
    # domains must be at least the number of classes
    domains: 10
    # alpha must be positive
    alpha: 0.5
    # max cond must be positive. If it is too small, it is unlikely that a valid matrix could be generated.
    max_condition_number: 4
    class_prior_seed: 4
  approaches:
  # options: ['ddfa', 'ddfa_scan']. If DDFA_SCAN is selected, the SCAN baseline will also be computed.
  - 'ddfa'
  - 'ddfa_scan'
- dataset_settings:
    dataset: 'cifar10'
    dataset_split_seed: 7636
  class_prior_generation:
    # domains must be at least the number of classes
    domains: 15
    # alpha must be positive
    alpha: 3
    # max cond must be positive. If it is too small, it is unlikely that a valid matrix could be generated.
    max_condition_number: 4
    class_prior_seed: 4
  approaches:
  # options: ['ddfa', 'ddfa_scan']. If DDFA_SCAN is selected, the SCAN baseline will also be computed.
  - 'ddfa'
  - 'ddfa_scan'
- dataset_settings:
    dataset: 'cifar10'
    dataset_split_seed: 25
  class_prior_generation:
    # domains must be at least the number of classes
    domains: 15
    # alpha must be positive
    alpha: 10
    # max cond must be positive. If it is too small, it is unlikely that a valid matrix could be generated.
    max_condition_number: 8
    class_prior_seed: 23
  approaches:
  # options: ['ddfa', 'ddfa_scan']. If DDFA_SCAN is selected, the SCAN baseline will also be computed.
  - 'ddfa'
  - 'ddfa_scan'

# In our paper, we conducted experiments using the following setups:
# cifar10:
#     Domain counts: [10,15,20,25]
#     Pairs of (alpha, max_condition_number): [(0.5, 4), (3, 4), (10, 8)]
#     Approaches: ['ddfa', 'ddfa_scan']
# cifar20:
#     Domain counts: [20,25,30]
#     Pairs of (alpha, max_condition_number): [(0.5, 8), (3, 12), (10, 20)]
#     Approaches: ['ddfa', 'ddfa_scan']
# imagenet:
#     Domain counts: [50,60]
#     Pairs of (alpha, max_condition_number): [(0.5, 200), (3, 205), (10, 210)]
#     Approaches: ['ddfa_scan']
# fg2:
#     Domain counts: [2, 3, 5, 7, 10]
#     Pairs of (alpha, max_condition_number): [(0.5, 3), (3, 5), (10, 7)]
#     Approaches: ['ddfa_scan']
# fg28:
#     Domain counts: [28, 37, 42, 47]
#     Pairs of (alpha, max_condition_number): [(0.5, 12), (3, 20), (10, 28)]
#     Approaches: ['ddfa_scan']