def test_all(self):
x = np.array([])
assert_array_equal(np.array([]), dhwt.transform(x)[0])
assert_array_equal(np.array([]), dhwt.transform(x)[1])
assert_array_equal(x, dhwt.inverse(*dhwt.transform(x)))
x = np.array([1., 1])
assert_array_equal(np.array([1.]), dhwt.transform(x)[0])
assert_array_equal(np.array([0.]), dhwt.transform(x)[1])
assert_array_equal(x, dhwt.inverse(*dhwt.transform(x)))
x = np.array([1., 2, 3, 0])
assert_array_equal(np.array([1.5, 1.5]), dhwt.transform(x)[0])
assert_array_equal(np.array([-.5, 1.5]), dhwt.transform(x)[1])
assert_array_equal(x, dhwt.inverse(*dhwt.transform(x)))
x = np.array([1., 2, 3, 0, 7])
assert_array_equal(np.array([1.5, 1.5, 3.5]), dhwt.transform(x)[0])
assert_array_equal(np.array([-.5, 1.5, 3.5]), dhwt.transform(x)[1])
assert_array_equal(x, dhwt.inverse(*dhwt.transform(x)))
x = np.array([6., 12, 15, 15, 14, 12, 120, 116])
assert_array_equal(np.array([9., 15, 13, 118]), dhwt.transform(x)[0])
assert_array_equal(np.array([-3, 0, 1, 2]), dhwt.transform(x)[1])
assert_array_equal(x, dhwt.inverse(*dhwt.transform(x)))
x = np.array([6., 12, 15, 15, 14, 12, 120, 116, 2])
assert_array_equal(np.array([9., 15, 13, 118, 1]), dhwt.transform(x)[0])
assert_array_equal(np.array([-3, 0, 1, 2, 1]), dhwt.transform(x)[1])
assert_array_equal(x, dhwt.inverse(*dhwt.transform(x)))
def inc_ksc(tseries, num_clusters, n_iters=-1, num_wavelets=2):
'''
Given the number `num_wavelets`, this method will compute subsequent
Discrete Harr Wavelet Transforms of the time series to be clustered. At
each transform the number of points of the time series is decreased, thus
we say that we are viewing the time series at a higher resolution.
Clustering will begin at the highest resolution (last transform), and the
results from the previous resolution is used to initialized the current one.
Only the highest resolution is initialized randomly. This technique can
improve the run-time of the KSC algorithm, since it is faster to cluster
at higher resolutions (less data points), being for subsequent resolutions
the centroids from the previous resolution already a close approximation of
the actual centroid. See [1] for details.
Please refer to the documentation of `_base_ksc` for a detailed summary
of the KSC algorithm.
Arguments
---------
tseries: a matrix of shape (number of time series, size of each series)
The time series to cluster
n_iters: int
The number of iterations which the algorithm will run
num_wavelets: int
The number of wavelets to use
Returns
-------
centroids: a matrix of shape (num. of clusters, size of time series)
The final centroids found by the algorithm
assign: an array of num. series size
The cluster id which each time series belongs to
best_shift: an array of num. series size
The amount shift amount performed for each time series
cent_dists: a matrix of shape (num. centroids, num. series)
The distance of each centroid to each time series
References
----------
.. [1] J. Yang and J. Leskovec,
"Patterns of Temporal Variation in Online Media" - WSDM'11
http://dl.acm.org/citation.cfm?id=1935863
'''
dhw_series = []
dhw_series.append(tseries)
previous = tseries
for _ in xrange(num_wavelets):
new_series = []
for j in xrange(tseries.shape[0]):
wave = transform(previous[j])[0]
new_series.append(wave)
previous = np.array(new_series)
dhw_series.append(previous)
assign = np.random.randint(0, num_clusters, tseries.shape[0])
cents = None
series_shift = None
for dhw in reversed(dhw_series):
cents = _compute_centroids(dhw, assign, num_clusters, series_shift)
cents, assign, series_shift, dists = _base_ksc(dhw, cents, n_iters)
return cents, assign, series_shift, dists
def inc_ksc(tseries, num_clusters, n_iters=-1, num_wavelets=2):
'''
Given the number `num_wavelets`, this method will compute subsequent
Discrete Harr Wavelet Transforms of the time series to be clustered. At
each transform the number of points of the time series is decreased, thus
we say that we are viewing the time series at a higher resolution.
Clustering will begin at the highest resolution (last transform), and the
results from the previous resolution is used to initialized the current one.
Only the highest resolution is initialized randomly. This technique can
improve the run-time of the KSC algorithm, since it is faster to cluster
at higher resolutions (less data points), being for subsequent resolutions
the centroids from the previous resolution already a close approximation of
the actual centroid. See [1] for details.
Please refer to the documentation of `_base_ksc` for a detailed summary
of the KSC algorithm.
Arguments
---------
tseries: a matrix of shape (number of time series, size of each series)
The time series to cluster
n_iters: int
The number of iterations which the algorithm will run
num_wavelets: int
The number of wavelets to use
Returns
-------
centroids: a matrix of shape (num. of clusters, size of time series)
The final centroids found by the algorithm
assign: an array of num. series size
The cluster id which each time series belongs to
best_shift: an array of num. series size
The amount shift amount performed for each time series
cent_dists: a matrix of shape (num. centroids, num. series)
The distance of each centroid to each time series
References
----------
.. [1] J. Yang and J. Leskovec,
"Patterns of Temporal Variation in Online Media" - WSDM'11
http://dl.acm.org/citation.cfm?id=1935863
'''
dhw_series = []
dhw_series.append(tseries)
previous = tseries
for _ in range(num_wavelets):
new_series = []
for j in range(tseries.shape[0]):
wave = transform(previous[j])[0]
new_series.append(wave)
previous = np.array(new_series)
dhw_series.append(previous)
assign = np.random.randint(0, num_clusters, tseries.shape[0])
cents = None
series_shift = None
for dhw in reversed(dhw_series):
cents = _compute_centroids(dhw, assign, num_clusters, series_shift)
cents, assign, series_shift, dists = _base_ksc(dhw, cents, n_iters)
return cents, assign, series_shift, dists
