def test_make_analysis_compressed_sensing_problem():
m = 10
N, n = 30, 20
l = 15
numdata = 10
assert_raises(ValueError, make_analysis_compressed_sensing_problem, m, n, N, l, numdata, "randn", "somethingwrong")
P = numpy.random.randn(m,n)
assert_raises(ValueError, make_analysis_compressed_sensing_problem, m+1, n, N, l, numdata, "randn", P)
measurements, acqumatrix, data, operator, gamma, cosupport = \
make_analysis_compressed_sensing_problem(m, n, N, l, numdata, "randn", "randn")
assert_equal(measurements.shape, (m,numdata))
assert_equal(acqumatrix.shape, (m,n))
assert_array_equal(measurements, numpy.dot(acqumatrix, data))
def run(self):
self.avgerr = dict()
for solver in self.solvers:
self.avgerr[solver] = np.zeros(shape=(len(self.deltas), len(self.rhos)))
for idelta, delta in enumerate(self.deltas):
for irho, rho in enumerate(self.rhos):
m = int(round(self.signaldim * delta, 0)) # delta = m/n
l = self.signaldim - int(round(m * rho, 0)) # rho = (n-l)/m
#TODO: make operator only once
measurements, acqumatrix, realdata, operator, realgamma, realcosupport = \
gen.make_analysis_compressed_sensing_problem(
m, self.signaldim, self.dictdim, l, self.numdata, "randn", "randn")
for solver in self.solvers:
data = solver.solve(measurements, acqumatrix, operator)
errors = data - realdata
for i in range(errors.shape[1]):
errors[:, i] = errors[:, i] / np.linalg.norm(realdata[:, i])
self.avgerr[solver][idelta, irho] = np.mean(np.sqrt(sum(errors**2, 0)))
def tuplewrap_make_analysis_compressed_sensing_problem(tuple_data):
return gen.make_analysis_compressed_sensing_problem(*tuple_data)
import numpy as np
from sklearn.utils.testing import assert_raises
from sklearn.utils.testing import assert_true
from sklearn.utils.testing import assert_equal
from sklearn.utils.testing import assert_allclose
from generate import make_analysis_compressed_sensing_problem
from analysisl1min import AnalysisL1Min
m, N, n, l, numdata = 16, 25, 20, 18, 10
#rng = np.random.RandomState(47)
SolverClass = AnalysisL1Min
measurements, acqumatrix, data, operator, gamma, cosupport = make_analysis_compressed_sensing_problem(m,n,N,l, numdata, random_state=47)
tol = 1e-6
algorithms = ["nesta"]
def test_correct_shapes():
stopvals = [1e-6]
for algo in algorithms:
for stopval in stopvals:
yield subtest_correct_shapes, stopval, algo
def subtest_correct_shapes(stopval, algorithm):
solver = SolverClass(stopval = stopval, algorithm=algorithm)
# single vector
recdata = solver.solve(measurements[:,0], acqumatrix, operator)
assert_equal(recdata.shape, (n,))
