def _preprocess_sklearn(self, X, y=None, fit_time=False):
force_all_finite = self.kernel not in VARIABLE_LENGTH_METRICS
if y is None:
X = check_array(X,
allow_nd=True,
force_all_finite=force_all_finite)
else:
X, y = check_X_y(X,
y,
allow_nd=True,
force_all_finite=force_all_finite)
X = to_time_series_dataset(X)
if fit_time:
self._X_fit = X
if self.gamma == "auto":
self.gamma_ = gamma_soft_dtw(X)
else:
self.gamma_ = self.gamma
self.classes_ = numpy.unique(y)
else:
check_is_fitted(self, ['svm_estimator_', '_X_fit'])
X = check_dims(X,
X_fit_dims=self._X_fit.shape,
extend=True,
check_n_features_only=(self.kernel
in VARIABLE_LENGTH_METRICS))
if self.kernel in VARIABLE_LENGTH_METRICS:
assert self.kernel == "gak"
self.estimator_kernel_ = "precomputed"
if fit_time:
sklearn_X = cdist_gak(X,
sigma=numpy.sqrt(self.gamma_ / 2.),
n_jobs=self.n_jobs,
verbose=self.verbose)
else:
sklearn_X = cdist_gak(X,
self._X_fit,
sigma=numpy.sqrt(self.gamma_ / 2.),
n_jobs=self.n_jobs,
verbose=self.verbose)
else:
self.estimator_kernel_ = self.kernel
sklearn_X = to_sklearn_dataset(X)
if y is None:
return sklearn_X
else:
return sklearn_X, y
def merge_crossover(ind1, ind2):
"""Merge shapelets from one set with shapelets from the other"""
# Construct a pairwise similarity matrix using GAK
_all = list(ind1) + list(ind2)
similarity_matrix = cdist_gak(ind1, ind2, sigma=sigma_gak(_all))
# Iterate over shapelets in `ind1` and merge them with shapelets
# from `ind2`
for row_idx in range(similarity_matrix.shape[0]):
# Remove all elements equal to 1.0
mask = similarity_matrix[row_idx, :] != 1.0
non_equals = similarity_matrix[row_idx, :][mask]
if len(non_equals):
# Get the timeseries most similar to the one at row_idx
max_col_idx = np.argmax(non_equals)
ts1 = list(ind1[row_idx]).copy()
ts2 = list(ind2[max_col_idx]).copy()
# Merge them and remove nans
ind1[row_idx] = euclidean_barycenter([ts1, ts2])
ind1[row_idx] = ind1[row_idx][~np.isnan(ind1[row_idx])]
# Apply the same for the elements in ind2
for col_idx in range(similarity_matrix.shape[1]):
mask = similarity_matrix[:, col_idx] != 1.0
non_equals = similarity_matrix[:, col_idx][mask]
if len(non_equals):
max_row_idx = np.argmax(non_equals)
ts1 = list(ind1[max_row_idx]).copy()
ts2 = list(ind2[col_idx]).copy()
ind2[col_idx] = euclidean_barycenter([ts1, ts2])
ind2[col_idx] = ind2[col_idx][~np.isnan(ind2[col_idx])]
return ind1, ind2
def sparse_gak(X, X_fit):
if X_fit is X:
return cdist_gak(X.reshape((-1, sz, d)), None, sigma=numpy.sqrt(gamma / 2.))
if slice_support_vectors is not None:
# slice out support vectors
sliced_X_fit = X_fit[slice_support_vectors]
gak_sim_dense = cdist_gak(X.reshape((-1, sz, d)), sliced_X_fit.reshape((-1, sz, d)), sigma=numpy.sqrt(gamma / 2.))
# act like nothing has happend ...
gak_sim = numpy.empty((len(X), len(X_fit)))
gak_sim[:, slice_support_vectors] = gak_sim_dense
return gak_sim
return cdist_gak(X.reshape((-1, sz, d)), X_fit.reshape((-1, sz, d)), sigma=numpy.sqrt(gamma / 2.))
def _kernel_func_gak(self, x, y):
g = self.estimator.gamma
sz = self.estimator.sz
d = self.estimator.d
if g == "auto":
g = 1.0
return cdist_gak(x.reshape((-1, sz, d)),
y.reshape((-1, sz, d)),
sigma=np.sqrt(g / 2.0))
def _preprocess_sklearn(self, X, y=None, fit_time=False):
force_all_finite = self.kernel not in VARIABLE_LENGTH_METRICS
if y is None:
X = check_array(X,
allow_nd=True,
force_all_finite=force_all_finite)
else:
X, y = check_X_y(X,
y,
allow_nd=True,
force_all_finite=force_all_finite)
X = check_dims(X, X_fit=None)
X = to_time_series_dataset(X)
if fit_time:
self._X_fit = X
self.gamma_ = gamma_soft_dtw(X)
self.classes_ = numpy.unique(y)
if self.kernel in VARIABLE_LENGTH_METRICS:
assert self.kernel == "gak"
self.estimator_kernel_ = "precomputed"
if fit_time:
sklearn_X = cdist_gak(X,
sigma=numpy.sqrt(self.gamma_ / 2.),
n_jobs=self.n_jobs,
verbose=self.verbose)
else:
sklearn_X = cdist_gak(X,
self._X_fit,
sigma=numpy.sqrt(self.gamma_ / 2.),
n_jobs=self.n_jobs,
verbose=self.verbose)
else:
self.estimator_kernel_ = self.kernel
sklearn_X = _prepare_ts_datasets_sklearn(X)
if y is None:
return sklearn_X
else:
return sklearn_X, y
def _get_kernel(self, X, Y=None):
kernel_params = self._get_kernel_params()
if self.kernel == "gak":
return cdist_gak(X, Y, n_jobs=self.n_jobs, verbose=self.verbose,
**kernel_params)
else:
X_sklearn = to_sklearn_dataset(X)
if Y is not None:
Y_sklearn = to_sklearn_dataset(Y)
else:
Y_sklearn = Y
return pairwise_kernels(X_sklearn, Y_sklearn, metric=self.kernel,
n_jobs=self.n_jobs, **kernel_params)
def test_gamma_value_svm():
n, sz, d = 5, 10, 3
rng = np.random.RandomState(0)
time_series = rng.randn(n, sz, d)
labels = rng.randint(low=0, high=2, size=n)
gamma = 10.
for ModelClass in [TimeSeriesSVC, TimeSeriesSVR]:
gak_model = ModelClass(kernel="gak", gamma=gamma)
sklearn_X, _ = gak_model._preprocess_sklearn(time_series,
labels,
fit_time=True)
cdist_mat = cdist_gak(time_series, sigma=np.sqrt(gamma / 2.))
np.testing.assert_allclose(sklearn_X, cdist_mat)
def _get_kernel(self, X, Y=None):
return cdist_gak(X, Y, sigma=self.sigma)
def _get_kernel(self, X, Y=None):
return cdist_gak(X, Y, sigma=self.sigma, n_jobs=self.n_jobs,
verbose=self.verbose)
def _kernel_func_gak(sz, d, gamma):
if gamma == "auto":
gamma = 1.
return lambda x, y: cdist_gak(x.reshape((-1, sz, d)),
y.reshape((-1, sz, d)),
sigma=numpy.sqrt(gamma / 2.))
