def __init__(self, dataset_path, sensor_idxs, action_idxs, max_laps, n_out, out_activation_f,
regression=True):
x_train, y_train, x_valid, y_valid, scaler = read_logs(dataset_path, sensor_idxs, action_idxs, max_laps,
noise=False, shuffled=True,
scale=True, valid_prop=.2)
# indices must be integers for classification
if regression:
dtype = theano.config.floatX
self.y_train = y_train.astype(dtype=theano.config.floatX)
self.y_valid = y_valid.astype(dtype=theano.config.floatX)
else:
dtype = int
self.y_train = y_train.astype(dtype=dtype)+1
self.y_valid = y_valid.astype(dtype=dtype)+1
self.x_train = x_train.astype(dtype=theano.config.floatX)
self.x_valid = x_valid.astype(dtype=theano.config.floatX)
self.sensor_idxs = sensor_idxs
self.action_idxs = action_idxs
self.max_laps = max_laps
self.scaler = scaler
self.network = None
self.n_out = n_out
self.out_activation_f = out_activation_f
self.out_activation_f_name = get_activation_function_name(out_activation_f)
self.regression = regression
def __init__(self, dataset_path, sensor_idxs, action_idxs, max_laps, n_hidden, n_out,
h0_activation_f, out_activation_f):
x_train, y_train, x_valid, y_valid, scaler = read_logs(dataset_path, sensor_idxs, action_idxs, max_laps,
noise=False, shuffled=True,
scale=True, valid_prop=.2)
self.x_train = x_train.astype(dtype=theano.config.floatX)
self.y_train = y_train.astype(dtype=theano.config.floatX)
self.x_valid = x_valid.astype(dtype=theano.config.floatX)
self.y_valid = y_valid.astype(dtype=theano.config.floatX)
self.sensor_idxs = sensor_idxs
self.action_idxs = action_idxs
self.max_laps = max_laps
self.scaler = scaler
self.network = None
self.n_hidden = n_hidden
self.n_out = n_out
self.h0_activation_f = h0_activation_f
self.h0_activation_f_name = get_activation_function_name(h0_activation_f)
self.out_activation_f = out_activation_f
self.out_activation_f_name = get_activation_function_name(out_activation_f)
