1 # File used to store hyperparameters.
3 # Stateless RNN, batch_size declared at fit step
10 dense_layers: 1 # hidden dense layers AFTER recurrent layers and BEFORE final output cell
11 dense_units: 5 # number of units for hidden dense layers
12 activation: ['linear', 'linear']
14 dropout: [0.2, 0.2] # NOTE: length must match total number of layers, default is 1 hidden recurrent layer and 1 dense output layer
15 recurrent_dropout: 0.2 # Length must match number of recurrent layers
16 reset_states: True # reset hidden states after training epoch, triggers reset_states() via callbacks
19 clipvalue: 10.0 # gradient clipping param, gradient can't exceed this value
20 phys_initialize: False # physics initialization
22 verbose_weights: True # Prints out hashs of weights for tracking reproducibility
23 verbose_fit: False # Prints out all training epochs, makes computation much slower
24 # features_list: ['Ed', 'Ew', 'solar', 'wind', 'rain']
25 features_list: ['Ed', 'Ew', 'rain']
27 scaler: 'minmax' # One of methods in scalers dictionary in moisture_rnn.py
31 # NOTE: as of 16-6-24 only param difference with simple rnn is recurrent_activation param
40 activation: ['linear', 'linear']
41 recurrent_activation: 'sigmoid'
43 dropout: [0.2, 0.2] # NOTE: length must match total number of layers, default is 1 hidden recurrent layer and 1 dense output layer
44 recurrent_dropout: 0.2 # Length must match number of recurrent layers
45 reset_states: True # reset hidden states after training epoch, triggers reset_states() via callbacks
48 clipvalue: 1.0 # gradient clipping param, gradient can't exceed this value
49 phys_initialize: False # physics initialization
51 verbose_weights: True # Prints out hashs of weights for tracking reproducibility
52 verbose_fit: False # Prints out all training epochs, makes computation much slower
53 features_list: ['Ed', 'Ew', 'rain']
55 scaler: 'minmax' # One of methods in scalers dictionary in moisture_rnn.py
59 # Lives in moisture_rnn now
60 # physics_initializer:
61 # DeltaE: [0,-1] # bias correction
62 # T1: 0.1 # 1/fuel class (10)
63 # fm_raise_vs_rain: 0.2 # fm increase per mm rain
66 # Param sets for reproducibility
77 activation: ['linear', 'linear']
80 recurrent_dropout: 0.2
85 phys_initialize: False
89 features_list: ['Ed', 'Ew', 'solar', 'wind', 'rain']
103 # dense_layers: 0 # hidden dense layers AFTER recurrent layers and BEFORE final output cell
105 # activation: ['linear', 'linear']
106 # centering: [0.0,0.0]
107 # dropout: [0.0, 0.0] # NOTE: length must match total number of layers, default is 1 hidden recurrent layer and 1 dense output layer
108 # recurrent_dropout: 0.0 # Length must match number of recurrent layers
109 # reset_states: True # reset hidden states after training epoch, triggers reset_states() via callbacks
111 # learning_rate: 0.001
112 # phys_initialize: False # physics initialization
114 # verbose_weights: True # Prints out hashs of weights for tracking reproducibility
115 # verbose_fit: False # Prints out all training epochs, makes computation much slower
116 # features_list: ['Ed', 'Ew', 'rain']
118 # scaler: 'reproducibility'