diff options
author | A. Unique TensorFlower <gardener@tensorflow.org> | 2018-09-11 17:39:28 -0700 |
---|---|---|
committer | TensorFlower Gardener <gardener@tensorflow.org> | 2018-09-11 17:43:39 -0700 |
commit | f4de1e737c914618e6fcefac5918fe73945ef9fb (patch) | |
tree | 9d9b57536d633da3b1e69c9f39f1c36c631c5a9c /tensorflow/contrib/timeseries | |
parent | e50233ce00d6010801934b9ac02f1d57de415672 (diff) |
Rename "_periods" private property in ARModel with "_periodicities" to make it more accurate.
PiperOrigin-RevId: 212555968
Diffstat (limited to 'tensorflow/contrib/timeseries')
-rw-r--r-- | tensorflow/contrib/timeseries/python/timeseries/ar_model.py | 16 |
1 files changed, 8 insertions, 8 deletions
diff --git a/tensorflow/contrib/timeseries/python/timeseries/ar_model.py b/tensorflow/contrib/timeseries/python/timeseries/ar_model.py index d808945334..1d27fffc62 100644 --- a/tensorflow/contrib/timeseries/python/timeseries/ar_model.py +++ b/tensorflow/contrib/timeseries/python/timeseries/ar_model.py @@ -264,10 +264,10 @@ class ARModel(model.TimeSeriesModel): elif (not isinstance(periodicities, list) and not isinstance(periodicities, tuple)): periodicities = [periodicities] - self._periods = [int(p) for p in periodicities] - for p in self._periods: + self._periodicities = [int(p) for p in periodicities] + for p in self._periodicities: assert p > 0 - assert len(self._periods) or self.input_window_size + assert len(self._periodicities) or self.input_window_size assert output_window_size > 0 def initialize_graph(self, input_statistics=None): @@ -364,9 +364,9 @@ class ARModel(model.TimeSeriesModel): input_feature_size = 0 output_window_features = [] output_feature_size = 0 - if self._periods: + if self._periodicities: _, time_features = self._compute_time_features(times) - num_time_features = self._buckets * len(self._periods) + num_time_features = self._buckets * len(self._periodicities) time_features = array_ops.reshape( time_features, [batch_size, @@ -849,12 +849,12 @@ class ARModel(model.TimeSeriesModel): def _compute_time_features(self, time): """Compute some features on the time value.""" batch_size = array_ops.shape(time)[0] - num_periods = len(self._periods) + num_periods = len(self._periodicities) # Reshape to 3D. periods = constant_op.constant( - self._periods, shape=[1, 1, num_periods, 1], dtype=time.dtype) + self._periodicities, shape=[1, 1, num_periods, 1], dtype=time.dtype) time = array_ops.reshape(time, [batch_size, -1, 1, 1]) - window_offset = time / self._periods + window_offset = time / self._periodicities # Cast to appropriate type and scale to [0, 1) range mod = (math_ops.cast(time % periods, self.dtype) * self._buckets / math_ops.cast(periods, self.dtype)) |