def test_TimeSeriesGenerator_doesnt_miss_any_sample(self):
x = np.array([[i] for i in range(10)])
for length in range(3, 10):
g = sequence.TimeseriesGenerator(x, x, length=length, batch_size=1)
expected = max(0, len(x) - length)
actual = len(g)
self.assertEqual(expected, actual)
if len(g) > 0:
# All elements in range(length, 10) should be used as current
# step
expected = np.arange(length, 10).reshape(-1, 1)
y = np.concatenate([g[ix][1] for ix in range(len(g))], axis=0)
self.assertAllClose(y, expected)
x = np.array([[i] for i in range(23)])
strides = (1, 1, 5, 7, 3, 5, 3)
lengths = (3, 3, 4, 3, 1, 3, 7)
batch_sizes = (6, 6, 6, 5, 6, 6, 6)
shuffles = (False, True, True, False, False, False, False)
for stride, length, batch_size, shuffle in zip(strides, lengths,
batch_sizes, shuffles):
g = sequence.TimeseriesGenerator(
x,
x,
length=length,
sampling_rate=1,
stride=stride,
start_index=0,
end_index=None,
shuffle=shuffle,
reverse=False,
batch_size=batch_size,
)
if shuffle:
# all batches have the same size when shuffle is True.
expected_sequences = (math.ceil(
(23 - length) / float(batch_size * stride)) * batch_size)
else:
# last batch will be different if `(samples - length) / stride`
# is not a multiple of `batch_size`.
expected_sequences = math.ceil((23 - length) / float(stride))
expected_batches = math.ceil(expected_sequences /
float(batch_size))
y = [g[ix][1] for ix in range(len(g))]
actual_sequences = sum(len(y_) for y_ in y)
actual_batches = len(y)
self.assertEqual(expected_sequences, actual_sequences)
self.assertEqual(expected_batches, actual_batches)
def get_time_series(data, labels, TIME_STEPS=10, START_INDEX=0):
try:
data_gen = sequence.TimeseriesGenerator(data,
labels,
length=TIME_STEPS,
sampling_rate=1,
batch_size=1,
start_index=int(TIME_STEPS /
2))
X = np.array([data_gen[i][0][0] for i in range(len(data_gen))])
Y = np.array([data_gen[i][1][0] for i in range(len(data_gen))])
except:
def TimeseriesGenerator(data, targets, length=10, start_index=0):
i = start_index
X, Y = [], []
while i < len(data) - length:
X.append(data[i:i + length])
# Y.append(targets[i:i+length])
Y.append(targets[i])
i += 1
X = np.array(X)
Y = np.array(Y)
return X, Y
X, Y = TimeseriesGenerator(data,
labels,
length=TIME_STEPS,
start_index=START_INDEX)
return X, Y
def test_TimeseriesGenerator(self):
data = np.array([[i] for i in range(50)])
targets = np.array([[i] for i in range(50)])
data_gen = preprocessing_sequence.TimeseriesGenerator(data,
targets,
length=10,
sampling_rate=2,
batch_size=2)
self.assertEqual(len(data_gen), 20)
self.assertAllClose(
data_gen[0][0],
np.array([[[0], [2], [4], [6], [8]], [[1], [3], [5], [7], [9]]]))
self.assertAllClose(data_gen[0][1], np.array([[10], [11]]))
self.assertAllClose(
data_gen[1][0],
np.array([[[2], [4], [6], [8], [10]], [[3], [5], [7], [9], [11]]]))
self.assertAllClose(data_gen[1][1], np.array([[12], [13]]))
data_gen = preprocessing_sequence.TimeseriesGenerator(data,
targets,
length=10,
sampling_rate=2,
reverse=True,
batch_size=2)
self.assertEqual(len(data_gen), 20)
self.assertAllClose(
data_gen[0][0],
np.array([[[8], [6], [4], [2], [0]], [[9], [7], [5], [3], [1]]]))
self.assertAllClose(data_gen[0][1], np.array([[10], [11]]))
data_gen = preprocessing_sequence.TimeseriesGenerator(data,
targets,
length=10,
sampling_rate=2,
shuffle=True,
batch_size=1)
batch = data_gen[0]
r = batch[1][0][0]
self.assertAllClose(
batch[0], np.array([[[r - 10], [r - 8], [r - 6], [r - 4],
[r - 2]]]))
self.assertAllClose(batch[1], np.array([
[r],
]))
data_gen = preprocessing_sequence.TimeseriesGenerator(data,
targets,
length=10,
sampling_rate=2,
stride=2,
batch_size=2)
self.assertEqual(len(data_gen), 10)
self.assertAllClose(
data_gen[1][0],
np.array([[[4], [6], [8], [10], [12]], [[6], [8], [10], [12],
[14]]]))
self.assertAllClose(data_gen[1][1], np.array([[14], [16]]))
data_gen = preprocessing_sequence.TimeseriesGenerator(data,
targets,
length=10,
sampling_rate=2,
start_index=10,
end_index=30,
batch_size=2)
self.assertEqual(len(data_gen), 6)
self.assertAllClose(
data_gen[0][0],
np.array([[[10], [12], [14], [16], [18]],
[[11], [13], [15], [17], [19]]]))
self.assertAllClose(data_gen[0][1], np.array([[20], [21]]))
data = np.array(
[np.random.random_sample((1, 2, 3, 4)) for i in range(50)])
targets = np.array(
[np.random.random_sample((3, 2, 1)) for i in range(50)])
data_gen = preprocessing_sequence.TimeseriesGenerator(data,
targets,
length=10,
sampling_rate=2,
start_index=10,
end_index=30,
batch_size=2)
self.assertEqual(len(data_gen), 6)
self.assertAllClose(
data_gen[0][0],
np.array([np.array(data[10:19:2]),
np.array(data[11:20:2])]))
self.assertAllClose(data_gen[0][1],
np.array([targets[20], targets[21]]))
with self.assertRaises(ValueError) as context:
preprocessing_sequence.TimeseriesGenerator(data,
targets,
length=50)
error = str(context.exception)
self.assertIn('`start_index+length=50 > end_index=49` is disallowed',
error)
def test_TimeseriesGenerator(self):
data = np.array([[i] for i in range(50)])
targets = np.array([[i] for i in range(50)])
data_gen = sequence.TimeseriesGenerator(data,
targets,
length=10,
sampling_rate=2,
batch_size=2)
self.assertLen(data_gen, 20)
self.assertAllClose(
data_gen[0][0],
np.array([[[0], [2], [4], [6], [8]], [[1], [3], [5], [7], [9]]]),
)
self.assertAllClose(data_gen[0][1], np.array([[10], [11]]))
self.assertAllClose(
data_gen[1][0],
np.array([[[2], [4], [6], [8], [10]], [[3], [5], [7], [9], [11]]]),
)
self.assertAllClose(data_gen[1][1], np.array([[12], [13]]))
data_gen = sequence.TimeseriesGenerator(
data,
targets,
length=10,
sampling_rate=2,
reverse=True,
batch_size=2,
)
self.assertLen(data_gen, 20)
self.assertAllClose(
data_gen[0][0],
np.array([[[8], [6], [4], [2], [0]], [[9], [7], [5], [3], [1]]]),
)
self.assertAllClose(data_gen[0][1], np.array([[10], [11]]))
data_gen = sequence.TimeseriesGenerator(
data,
targets,
length=10,
sampling_rate=2,
shuffle=True,
batch_size=1,
)
batch = data_gen[0]
r = batch[1][0][0]
self.assertAllClose(
batch[0], np.array([[[r - 10], [r - 8], [r - 6], [r - 4],
[r - 2]]]))
self.assertAllClose(
batch[1],
np.array([
[r],
]),
)
data_gen = sequence.TimeseriesGenerator(data,
targets,
length=10,
sampling_rate=2,
stride=2,
batch_size=2)
self.assertLen(data_gen, 10)
self.assertAllClose(
data_gen[1][0],
np.array([[[4], [6], [8], [10], [12]], [[6], [8], [10], [12],
[14]]]),
)
self.assertAllClose(data_gen[1][1], np.array([[14], [16]]))
data_gen = sequence.TimeseriesGenerator(
data,
targets,
length=10,
sampling_rate=2,
start_index=10,
end_index=30,
batch_size=2,
)
self.assertLen(data_gen, 6)
self.assertAllClose(
data_gen[0][0],
np.array([[[10], [12], [14], [16], [18]],
[[11], [13], [15], [17], [19]]]),
)
self.assertAllClose(data_gen[0][1], np.array([[20], [21]]))
data = np.array(
[np.random.random_sample((1, 2, 3, 4)) for i in range(50)])
targets = np.array(
[np.random.random_sample((3, 2, 1)) for i in range(50)])
data_gen = sequence.TimeseriesGenerator(
data,
targets,
length=10,
sampling_rate=2,
start_index=10,
end_index=30,
batch_size=2,
)
self.assertLen(data_gen, 6)
self.assertAllClose(
data_gen[0][0],
np.array([np.array(data[10:19:2]),
np.array(data[11:20:2])]),
)
self.assertAllClose(data_gen[0][1],
np.array([targets[20], targets[21]]))
with self.assertRaisesRegex(
ValueError,
r"`start_index\+length=50 > end_index=49` is disallowed"):
sequence.TimeseriesGenerator(data, targets, length=50)