swafa/params.yaml at master · ayeright/swafa · 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
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
online_fa:
  gradient_optimiser: 'sgd'
  gradient_optimiser_kwargs:
    lr: 0.001
  gradient_warm_up_time_steps: 100
  em_warm_up_time_steps: 100
  experiments:
    - {observation_dim: 100, latent_dim: 10, spectrum_range: [1, 10], n_samples: [100, 200, 500, 1000, 2000, 5000, 10000, 20000, 50000, 100000]}
    - {observation_dim: 100, latent_dim: 10, spectrum_range: [1, 100], n_samples: [100, 200, 500, 1000, 2000, 5000, 10000, 20000, 50000, 100000]}
    - {observation_dim: 100, latent_dim: 10, spectrum_range: [1, 1000], n_samples: [100, 200, 500, 1000, 2000, 5000, 10000, 20000, 50000, 100000]}
    - {observation_dim: 100, latent_dim: 10, spectrum_range: [1, 10000], n_samples: [100, 200, 500, 1000, 2000, 5000, 10000, 20000, 50000, 100000]}
    - {observation_dim: 1000, latent_dim: 10, spectrum_range: [1, 10], n_samples: [100, 200, 500, 1000, 2000, 5000, 10000, 20000, 50000, 100000]}
    - {observation_dim: 1000, latent_dim: 10, spectrum_range: [1, 100], n_samples: [100, 200, 500, 1000, 2000, 5000, 10000, 20000, 50000, 100000]}
    - {observation_dim: 1000, latent_dim: 10, spectrum_range: [1, 1000], n_samples: [100, 200, 500, 1000, 2000, 5000, 10000, 20000, 50000, 100000]}
    - {observation_dim: 1000, latent_dim: 10, spectrum_range: [1, 10000], n_samples: [100, 200, 500, 1000, 2000, 5000, 10000, 20000, 50000, 100000]}
  n_trials: 10

online_fa_analysis:
  min_samples: 100

linear_regression_vi:
  testing: true
  datasets:
    boston_housing:
      latent_dim: 3
      n_gradients_per_update: 10
      optimiser: 'sgd'
      bias_optimiser_kwargs:
        lr: 0.001
      factors_optimiser_kwargs:
        lr: 0.001
      noise_optimiser_kwargs:
        lr: 0.001
      max_grad_norm: 10
      batch_size: 100
      n_epochs: 25000
    yacht_hydrodynamics:
      latent_dim: 3
      n_gradients_per_update: 10
      optimiser: 'sgd'
      bias_optimiser_kwargs:
        lr: 0.01
      factors_optimiser_kwargs:
        lr: 0.01
      noise_optimiser_kwargs:
        lr: 0.01
      max_grad_norm: 10
      batch_size: 100
      n_epochs: 45000
    concrete_strength:
      latent_dim: 3
      n_gradients_per_update: 10
      optimiser: 'sgd'
      bias_optimiser_kwargs:
        lr: 0.01
      factors_optimiser_kwargs:
        lr: 0.01
      noise_optimiser_kwargs:
        lr: 0.01
      max_grad_norm: 10
      batch_size: 100
      n_epochs: 20000
    energy_efficiency:
      latent_dim: 3
      n_gradients_per_update: 10
      optimiser: 'sgd'
      bias_optimiser_kwargs:
        lr: 0.01
      factors_optimiser_kwargs:
        lr: 0.01
      noise_optimiser_kwargs:
        lr: 0.01
      max_grad_norm: 10
      batch_size: 100
      n_epochs: 25000

neural_net_predictions:
  n_cv_folds: 5 # same as https://arxiv.org/pdf/1811.04504.pdf
  n_hyperparameter_trials: 30 # same as https://arxiv.org/pdf/1811.04504.pdf
  hidden_dims: [50] # same as https://arxiv.org/pdf/1811.04504.pdf
  hidden_activation_fn: 'relu' # same as https://arxiv.org/pdf/1811.04504.pdf
  data_split_random_seed: 1 # same as https://github.com/yaringal/DropoutUncertaintyExps/tree/master/UCI_Datasets
  test: true
  n_train_test_splits: 20 # same as https://arxiv.org/pdf/1811.04504.pdf
  train_fraction: 0.9  # same as https://arxiv.org/pdf/1811.04504.pdf
  datasets: # same as https://github.com/aaronpmishkin/SLANG/tree/master/code/python/libs/vi_lib/experiments/uci
    boston_housing:
      latent_dim: 1
      n_gradients_per_update: 4
      max_grad_norm: 10
      batch_size: 10
      n_epochs: 120
      learning_rate_range: [0.01, 0.02]
      prior_precision_range: [0.01, 10]
      noise_precision_range: [0.01, 1]
      n_bma_samples: 100
    yacht_hydrodynamics:
      latent_dim: 1
      n_gradients_per_update: 4
      max_grad_norm: 10
      batch_size: 10
      n_epochs: 120
      learning_rate_range: [0.01, 0.02]
      prior_precision_range: [0.01, 10]
      noise_precision_range: [0.01, 1]
      n_bma_samples: 100
    concrete_strength:
      latent_dim: 1
      n_gradients_per_update: 4
      max_grad_norm: 10
      batch_size: 10
      n_epochs: 120
      learning_rate_range: [0.01, 0.02]
      prior_precision_range: [0.01, 10]
      noise_precision_range: [0.01, 1]
      n_bma_samples: 100
    energy_efficiency:
      latent_dim: 1
      n_gradients_per_update: 4
      max_grad_norm: 10
      batch_size: 10
      n_epochs: 120
      learning_rate_range: [0.01, 0.02]
      prior_precision_range: [0.01, 10]
      noise_precision_range: [0.01, 1]
      n_bma_samples: 100