def test_rsvd_complex64(self):
m, n = 5, 4
a = np.array(np.random.randn(m, n) + 1j * np.random.randn(m, n),
np.complex64,
order='F')
a_gpu = gpuarray.to_gpu(a)
U, s, Vt = rlinalg.rsvd(a_gpu, k=n, p=0, q=2, method='standard')
assert np.allclose(a,
np.dot(U.get(), np.dot(np.diag(s.get()), Vt.get())),
atol_float32)
def test_rsvdf_float32(self):
m, n = 5, 4
a = np.array(np.random.randn(m, n), np.float32, order='F')
a_gpu = gpuarray.to_gpu(a)
U, s, Vt = rlinalg.rsvd(a_gpu, k=n, p=0, q=2, method='fast')
assert np.allclose(a, np.dot(U.get(), np.dot(np.diag(s.get()), Vt.get())), atol_float32)
import pycuda.gpuarray as gpuarray
import pycuda.autoinit
import numpy as np
from skcuda import linalg, rlinalg
import sys
import timeit
from datetime import datetime
from sklearn.decomposition import RandomizedPCA
linalg.init()
rlinalg.init()
a = np.load(
open('/if10/spk3rw/nytimesdata/matrix_' + str(sys.argv[1]) + '_docs.bin'))
a = a.astype(np.float32)
a_gpu = gpuarray.to_gpu(a.T)
t = datetime.now()
U, s, Vt = rlinalg.rsvd(a_gpu, k=50, method='standard')
print("GPU Time:")
print(datetime.now() - t)
print(U.shape, s.shape, Vt.shape)
X = a.T
pca = RandomizedPCA(n_components=50)
t = datetime.now()
pca.fit(X)
print("CPU Time:")
print(datetime.now() - t)
print(pca.explained_variance_ratio_)
def test_rsvdf_complex128(self):
m, n = 5, 4
a = np.array(np.random.randn(m, n) + 1j*np.random.randn(m, n), np.complex128, order='F')
a_gpu = gpuarray.to_gpu(a)
U, s, Vt = rlinalg.rsvd(a_gpu, k=n, p=0, q=2, method='fast')
assert np.allclose(a, np.dot(U.get(), np.dot(np.diag(s.get()), Vt.get())), atol_float64)
def test_rsvd_float64(self):
m, n = 5, 4
a = np.array(np.random.randn(m, n), np.float64, order='F')
a_gpu = gpuarray.to_gpu(a)
U, s, Vt = rlinalg.rsvd(a_gpu, k=n, p=0, q=2, method='standard')
assert np.allclose(a, np.dot(U.get(), np.dot(np.diag(s.get()), Vt.get())), atol_float64)
import pycuda.gpuarray as gpuarray
import pycuda.autoinit
import numpy as np
from skcuda import linalg, rlinalg
import sys
import timeit
from datetime import datetime
from sklearn.decomposition import RandomizedPCA
linalg.init()
rlinalg.init()
a = np.load(open('/if10/spk3rw/nytimesdata/matrix_'+str(sys.argv[1])+'_docs.bin'))
a = a.astype(np.float32)
a_gpu = gpuarray.to_gpu(a.T)
t = datetime.now()
U, s, Vt = rlinalg.rsvd(a_gpu, k=50, method='standard')
print("GPU Time:")
print(datetime.now()-t)
print(U.shape,s.shape,Vt.shape)
X = a.T
pca = RandomizedPCA(n_components=50)
t = datetime.now()
pca.fit(X)
print("CPU Time:")
print(datetime.now()-t)
print(pca.explained_variance_ratio_)
