'lambda': 0,
'af': 'relu', # 'relu', 'linear', 'sigmoid'
'bias': True,
'window_size': 30,
# train parameters -------------------------
'loss_function': 'mae', # 'mse' or 'mae'
'lr': 0.001,
'batch_size': 2560,
'shuffle': False,
'normalization':
'standard_positive', # 'min_max', 'standard', None
'epochs': 250
}
utrain.train_and_test(params)
ufig.plot_train_test(ids_attempt[-1], [3, 8, 13, 17],
lim_nr_samples=2000)
ufig.plot_multi_boxplots(ids=ids_attempt,
x='patient_id',
y='correlation',
hue='brain_state',
save_name=pre + 'corr' + post)
ufig.plot_multi_boxplots(ids=ids_attempt,
x='patient_id',
y='mae',
hue='brain_state',
save_name=pre + 'mae' + post)
ufig.plot_multi_boxplots(ids=ids_attempt,
x='patient_id',
y='mse',
hue='brain_state',
'patient_id': val[1],
'time_begin': val[2], # [hour, minute]
'duration': 10, # seconds
'brain_state': val[3],
'add_id': '(M)',
# model parameters ------------------------
'visible_size': 'all', # 'all' or scalar
'hidden_size': 0, # improve: portion
'lambda': 0,
'af': 'relu', # 'relu', 'linear', 'sigmoid'
'bias': True,
'window_size': 0,
'resample': 512,
# train parameters -------------------------
'loss_function': 'mae', # 'mse' or 'mae'
'lr': 0.0002,
'batch_size': 5,
'shuffle': False,
'normalization': 'standard_positive', # 'min_max', 'standard', None
'epochs': 23}
utrain.train_and_test(params)
ufig.plot_train_test(ids_all[-1], [1, 3, 5, 7])
ufig.plot_multi_boxplots(ids=ids_attempt, x='batch_size', y='correlation', hue='brain_state',
save_name=pre + 'corr' + post, ylim=(0, 1))
#ufig.plot_multi_boxplots(ids=ids_attempt, x='patient_id', y='mae', hue='brain_state', save_name=pre + 'mae' + post)
#ufig.plot_multi_boxplots(ids=ids_attempt, x='patient_id', y='mse', hue='brain_state', save_name=pre + 'mse' + post)
ufig.mean_weights(ids=ids_all, save_name=pre)