def test_mini_batch_correct_shapes():
rng = np.random.RandomState(0)
X = rng.randn(12, 10)
pca = MiniBatchSparsePCA(n_components=8, random_state=rng)
U = pca.fit_transform(X)
assert_equal(pca.components_.shape, (8, 10))
assert_equal(U.shape, (12, 8))
# test overcomplete decomposition
pca = MiniBatchSparsePCA(n_components=13, random_state=rng)
U = pca.fit_transform(X)
assert_equal(pca.components_.shape, (13, 10))
assert_equal(U.shape, (12, 13))
def test_mini_batch_correct_shapes():
rng = np.random.RandomState(0)
X = rng.randn(12, 10)
pca = MiniBatchSparsePCA(n_components=8, random_state=rng)
U = pca.fit_transform(X)
assert_equal(pca.components_.shape, (8, 10))
assert_equal(U.shape, (12, 8))
# test overcomplete decomposition
pca = MiniBatchSparsePCA(n_components=13, random_state=rng)
U = pca.fit_transform(X)
assert_equal(pca.components_.shape, (13, 10))
assert_equal(U.shape, (12, 13))
def decompose_minibatch_sparse_pca(X,
n_components,
alpha=0.8,
n_iter=100,
batch_size=3,
random_state=np.random.RandomState(42)):
minibatch_sparse_pca = MiniBatchSparsePCA(
n_components=n_components,
alpha=alpha,
n_iter=n_iter,
batch_size=batch_size,
random_state=random_state,
)
X_minibatch_sparse_pca = minibatch_sparse_pca.fit_transform(X)
return X_minibatch_sparse_pca
def cluster_sk_mini_batch_sparse_pca(content):
""" x """
_config = MiniBatchSparsePCA(n_components=content['n_components'],
alpha=content['alpha'],
ridge_alpha=content['ridge_alpha'],
n_iter=content['n_iter'],
callback=None,
batch_size=content['batch_size'],
verbose=0,
shuffle=content['shuffle'],
n_jobs=-1,
method=content['sk_method'],
random_state=None)
_result = _config.fit_transform(content['data'])
return httpWrapper(
json.dumps({
'result': _result.tolist(),
'components': _config.components_.tolist(),
'iter': _config.n_iter_
}))
if idx in select_word_idx_list: # print idx,len(fea) li.append(fea[idx]) ALL_FEA[i] = li NeedPCA = False if NeedPCA: print 'Len of ALL_FEA: ',len(ALL_FEA) print 'Start PCA ... ' # pca = KernelPCA(n_components = num_af_pca,) pca = MiniBatchSparsePCA(n_components = num_af_pca,n_jobs = 4,verbose = 1,batch_size = len(ALL_FEA)/10) new_all_fea = pca.fit_transform(np.array(ALL_FEA)) # from sklearn.discriminant_analysis import LinearDiscriminantAnalysis as LDA # pca = LDA(n_components = num_af_pca) # new_all_fea = pca.fit_transform(np.array(ALL_FEA)) ALL_FEA = new_all_fea print '\nFinish PCA ... ' allSongs = [] head = 0 tail = 0 for gen in range(sum(usedGenres)):
