def compare_to_previous():
d = cached_surrogate_data(fname='tsa_p2_d0_q0',
k=1,
n_real=60,
n_samples=150,
n_warm=100,
n_input=80)
summary = ktrain(d=d, epochs=50000)
ratio = summary['test_error'] / TSA_P2_D0_Q0_BEST['test_error']
summary['test_error_ratio'] = ratio
return summary
keras.metrics.mean_squared_error(y_test_hat[:, 0, 0], d['y_test'][:,
0]))
y_val_hat = model.predict(d['x_val'])
val_error = float(
keras.metrics.mean_squared_error(y_val_hat[:, 0, 0], d['y_val'][:, 0]))
y_train_hat = model.predict(d['x_train'])
train_error = float(
keras.metrics.mean_squared_error(y_train_hat[:, 0, 0],
d['y_train'][:, 0]))
summary = {
"train_error": train_error,
"val_error": val_error,
"test_error": test_error
}
return summary
if __name__ == '__main__':
n_inputs = 80
epochs = 500
d = cached_surrogate_data(fname='tsa_p2_d0_q0',
k=1,
n_real=60,
n_samples=150,
n_warm=100,
n_input=n_inputs)
best_model = tunertrain(d=d, n_input=n_inputs, epochs=epochs)
summary = summarize_model(d=d, model=best_model)
ratio = summary['test_error'] / SLUGGISH_MOVING_AVERAGE_BEST['test_error']
summary['test_error_ratio'] = ratio
pprint(summary)
def compare_to_previous(): d = cached_surrogate_data(fname='sluggish_moving_average', k=1, n_real=50, n_samples=150, n_warm = 100, n_input=80) summary = ktrain(d=d, epochs=1000) ratio = summary['test_error'] / SLUGGISH_MOVING_AVERAGE_BEST['test_error'] summary['test_error_ratio'] = ratio return summary
def compare_to_previous(): d = cached_surrogate_data(fname='thinking_slow_and_fast', k=1, n_real=50, n_samples=150, n_warm = 100, n_input=80) summary = ktrain(d=d, epochs=5000) ratio = summary['test_error'] / SLOW_AND_FAST_BEST['test_error'] summary['test_error_ratio'] = ratio return summary
keras.metrics.mean_squared_error(y_test_hat[:, 0, 0], d['y_test'][:,
0]))
y_val_hat = model.predict(d['x_val'])
val_error = float(
keras.metrics.mean_squared_error(y_val_hat[:, 0, 0], d['y_val'][:, 0]))
y_train_hat = model.predict(d['x_train'])
train_error = float(
keras.metrics.mean_squared_error(y_train_hat[:, 0, 0],
d['y_train'][:, 0]))
summary = {
"train_error": train_error,
"val_error": val_error,
"test_error": test_error
}
return summary
if __name__ == '__main__':
n_inputs = 80
epochs = 500
d = cached_surrogate_data(fname='sluggish_moving_average',
k=1,
n_real=50,
n_samples=150,
n_warm=100,
n_input=n_inputs)
best_model = tunertrain(d=d, n_input=n_inputs, epochs=epochs)
summary = summarize_model(d=d, model=best_model)
ratio = summary['test_error'] / SLUGGISH_MOVING_AVERAGE_BEST['test_error']
summary['test_error_ratio'] = ratio
pprint(summary)
keras.metrics.mean_squared_error(y_test_hat[:, 0, 0], d['y_test'][:,
0]))
y_val_hat = model.predict(d['x_val'])
val_error = float(
keras.metrics.mean_squared_error(y_val_hat[:, 0, 0], d['y_val'][:, 0]))
y_train_hat = model.predict(d['x_train'])
train_error = float(
keras.metrics.mean_squared_error(y_train_hat[:, 0, 0],
d['y_train'][:, 0]))
summary = {
"train_error": train_error,
"val_error": val_error,
"test_error": test_error
}
return summary
if __name__ == '__main__':
n_inputs = 80
epochs = 500
d = cached_surrogate_data(fname='slow_precision_ema_ensemble',
k=1,
n_real=80,
n_samples=150,
n_warm=100,
n_input=n_inputs)
best_model = tunertrain(d=d, n_input=n_inputs, epochs=epochs)
summary = summarize_model(d=d, model=best_model)
ratio = summary['test_error'] / SLUGGISH_MOVING_AVERAGE_BEST['test_error']
summary['test_error_ratio'] = ratio
pprint(summary)
overwrite=True,
max_epochs=2500)
tuner.search(d['x_train'], d['y_train'], epochs=epochs,
validation_data=(d['x_val'], d['y_val']),callbacks = [callback])
print(tuner.results_summary())
best_model = tuner.get_best_models()[0]
return best_model
def summarize_model(d, model):
y_test_hat = model.predict(d['x_test'])
test_error = float(keras.metrics.mean_squared_error(y_test_hat[:, 0, 0], d['y_test'][:, 0]))
y_val_hat = model.predict(d['x_val'])
val_error = float(keras.metrics.mean_squared_error(y_val_hat[:, 0, 0], d['y_val'][:, 0]))
y_train_hat = model.predict(d['x_train'])
train_error = float(keras.metrics.mean_squared_error(y_train_hat[:, 0, 0], d['y_train'][:, 0]))
summary = {"train_error": train_error,
"val_error": val_error,
"test_error": test_error}
return summary
if __name__=='__main__':
n_inputs = 80
epochs = 500
d = cached_surrogate_data(fname='thinking_slow_and_fast', k=1, n_real=50, n_samples=150, n_warm=100, n_input=n_inputs)
best_model = tunertrain(d=d,n_input=n_inputs, epochs=epochs)
summary = summarize_model(d=d, model=best_model)
ratio = summary['test_error'] / SLUGGISH_MOVING_AVERAGE_BEST['test_error']
summary['test_error_ratio'] = ratio
pprint(summary)