def setup(self):
filename = "$GAMMAPY_DATA/tests/unbundled/poisson_stats_image/counts.fits.gz"
image = Map.read(filename)
background = image.copy(data=np.ones(image.data.shape, dtype=float))
self.kernels = CWTKernels(n_scale=2,
min_scale=3.0,
step_scale=2.6,
old=False)
self.data = dict(image=image, background=background)
self.cwt = CWT(kernels=self.kernels,
significance_threshold=2.0,
keep_history=True)
def run_cwt():
data = make_poisson_data()
cwt_kernels = CWTKernels(n_scale=2,
min_scale=3.0,
step_scale=2.6,
old=False)
cwt = CWT(kernels=cwt_kernels,
significance_threshold=2.,
keep_history=True)
cwt_data = CWTData(counts=data['image'],
background=data['background'],
n_scale=cwt_kernels.n_scale)
cwt.analyze(data=cwt_data)
return cwt_data
def run_cwt():
data = make_poisson_data()
cwt_kernels = CWTKernels(n_scale=2,
min_scale=3.0,
step_scale=2.6,
old=False)
cwt = CWT(kernels=cwt_kernels,
significance_threshold=2.,
keep_history=True)
cwt_data = CWTData(counts=data['image'],
background=data['background'],
n_scale=cwt_kernels.n_scale)
cwt.analyze(data=cwt_data)
return cwt_data
def run_cwt(data):
cwt = CWT(nscales=2, min_scale=6.0, scale_step=1.3)
data['background'] = 1. * np.ones_like(data['background'], dtype=float)
cwt.set_data(data['image'], data['background'])
# TODO: run one step, try to understand what it does
# cwt.run_one_iteration(nsigma=100, nsigmap=4.0)
cwt.run_iteratively(nsigma=100, niter=5)
return cwt
# Let's start to analyse input data. Import Logging module to see how the algorithm works during data analysis.
# In[ ]:
from gammapy.detect import CWT
import logging
logger = logging.getLogger()
logger.setLevel(logging.INFO)
cwt = CWT(
kernels=cwt_kernels,
tol=TOL,
significance_threshold=SIGNIFICANCE_THRESHOLD,
significance_island_threshold=SIGNIFICANCE_ISLAND_THRESHOLD,
remove_isolated=REMOVE_ISOLATED,
keep_history=KEEP_HISTORY,
)
# In order to the algorithm was able to analyze source images, you need to convert them to a special format, i.e. create an CWTData object. Do this.
# In[ ]:
from gammapy.detect import CWTKernels, CWTData
cwt_data = CWTData(
counts=data["counts"], background=data["background"], n_scale=N_SCALE
)
class TestCWT:
"""Test CWT algorithm."""
def setup(self):
filename = "$GAMMAPY_DATA/tests/unbundled/poisson_stats_image/counts.fits.gz"
image = Map.read(filename)
background = image.copy(data=np.ones(image.data.shape, dtype=float))
self.kernels = CWTKernels(n_scale=2,
min_scale=3.0,
step_scale=2.6,
old=False)
self.data = dict(image=image, background=background)
self.cwt = CWT(kernels=self.kernels,
significance_threshold=2.0,
keep_history=True)
def test_execute_iteration(self):
cwt_data = CWTData(
counts=self.data["image"],
background=self.data["background"],
n_scale=self.kernels.n_scale,
)
self.cwt._execute_iteration(data=cwt_data)
residual = cwt_data.residual.data
assert_allclose(residual.var(), 1.10209961137)
assert_allclose(residual[100, 100], 4.60067024083)
assert_allclose(residual[10, 10], 0.0210108511546)
def test_transform(self):
cwt_data = CWTData(
counts=self.data["image"],
background=self.data["background"],
n_scale=self.kernels.n_scale,
)
self.cwt._transform(data=cwt_data)
transform_3d = cwt_data.transform_3d.data
assert_allclose(transform_3d[0, 100, 100], 0.0444647513236)
assert_allclose(transform_3d[0, 10, 10], -0.00133091756454)
assert_allclose(transform_3d[1, 100, 100], 0.00165322855919)
assert_allclose(transform_3d[1, 10, 10], -9.2715980927e-05)
error = cwt_data.error.data
assert_allclose(error[0, 100, 100], 0.000840670230257)
assert_allclose(error[0, 10, 10], 0.000810230288381)
assert_allclose(error[1, 100, 100], 4.78305652411e-05)
assert_allclose(error[1, 10, 10], 4.02498840476e-05)
approx = cwt_data.approx.data
assert_allclose(approx[100, 100], 0.353211779292)
assert_allclose(approx[10, 10], -0.0210108511546)
approx_bkg = cwt_data.approx_bkg.data
assert_allclose(approx_bkg[100, 100], 0.99988318386)
assert_allclose(approx_bkg[10, 10], 0.486747980289)
def test_compute_support(self):
cwt_data = CWTData(
counts=self.data["image"],
background=self.data["background"],
n_scale=self.kernels.n_scale,
)
self.cwt._transform(data=cwt_data)
self.cwt._compute_support(data=cwt_data)
support_3d = cwt_data.support_3d.data
assert_allclose(support_3d[0].sum(), 1095)
assert_allclose(support_3d[1].sum(), 2368)
def test_inverse_transform(self):
cwt_data = CWTData(
counts=self.data["image"],
background=self.data["background"],
n_scale=self.kernels.n_scale,
)
self.cwt._execute_iteration(data=cwt_data)
model = cwt_data.model.data
transform_2d = cwt_data.transform_2d.data
assert_allclose(model.sum(), 11.7236771527)
assert_allclose(transform_2d.sum(), 11.7236771527)
def test_all_cwt_iterations(self):
cwt_data = CWTData(
counts=self.data["image"],
background=self.data["background"],
n_scale=self.kernels.n_scale,
)
self.cwt.analyze(data=cwt_data)
transform_3d = cwt_data.transform_3d.data
assert_allclose(transform_3d[0, 100, 100], 0.0401320295446)
assert_allclose(transform_3d[0, 10, 10], -0.00117538066327)
assert_allclose(transform_3d[1, 100, 100], 0.00112861578719)
assert_allclose(transform_3d[1, 10, 10], -6.86491626269e-05)
error = cwt_data.error.data
assert_allclose(error[0, 100, 100], 0.00105275393856)
assert_allclose(error[0, 10, 10], 0.000801986157367)
assert_allclose(error[1, 100, 100], 5.54110995048e-05)
assert_allclose(error[1, 10, 10], 3.9892257794e-05)
approx = cwt_data.approx.data
assert_allclose(approx[100, 100], 0.323369470219)
assert_allclose(approx[10, 10], -0.0210240420041)
approx_bkg = cwt_data.approx_bkg.data
assert_allclose(approx_bkg[100, 100], 0.99988318386)
assert_allclose(approx_bkg[10, 10], 0.486747980289)
support_3d = cwt_data.support_3d.data
assert_allclose(support_3d[0].sum(), 1151)
assert_allclose(support_3d[1].sum(), 2368)
model = cwt_data.model.data
transform_2d = cwt_data.transform_2d.data
assert_allclose(model.sum(), 103.96476418)
assert_allclose(transform_2d.sum(), 9.91731463861)
class TestCWTData:
"""
Test CWTData class.
"""
def setup(self):
filename = "$GAMMAPY_DATA/tests/unbundled/poisson_stats_image/counts.fits.gz"
image = Map.read(filename)
background = image.copy(data=np.ones(image.data.shape, dtype=float))
self.kernels = CWTKernels(n_scale=2,
min_scale=3.0,
step_scale=2.6,
old=False)
self.data = dict(image=image, background=background)
self.cwt = CWT(kernels=self.kernels,
significance_threshold=2.0,
keep_history=True)
self.cwt_data = CWTData(counts=image,
background=background,
n_scale=self.kernels.n_scale)
self.cwt.analyze(data=self.cwt_data)
def test_images(self):
images = self.cwt_data.images()
assert_allclose(images["counts"].data[25, 25],
self.data["image"].data[25, 25])
assert_allclose(images["background"].data[36, 63],
self.data["background"].data[36, 63])
model_plus_approx = images["model_plus_approx"].data
assert_allclose(model_plus_approx[100, 100], 0.753205544726)
assert_allclose(model_plus_approx[10, 10], -0.0210240420041)
maximal = images["maximal"].data
assert_allclose(maximal[100, 100], 0.0401320295446)
assert_allclose(maximal[10, 10], 0.0)
support_2d = images["support_2d"].data
assert_allclose(support_2d.sum(), 2996)
def test_cube_metrics_info(self):
cubes = self.cwt_data.cubes()
name = "transform_3d"
cube = cubes[name].data
info = self.cwt_data._metrics_info(data=cube, name=name)
assert_equal(info["Shape"], "3D cube")
assert_allclose(info["Variance"], 3.24405547338e-06)
assert_allclose(info["Max value"], 0.041216412114)
def test_image_info(self):
t = self.cwt_data.image_info(name="residual")
assert_equal(t.colnames, ["Metrics", "Source"])
assert_equal(len(t), 7)
def test_cube_info(self):
t = self.cwt_data.cube_info(name="error")
assert_equal(t.colnames, ["Metrics", "Source"])
assert_equal(len(t), 7)
t = self.cwt_data.cube_info(name="error", per_scale=True)
assert_equal(t.colnames, ["Scale power", "Metrics", "Source"])
assert_equal(len(t), 14)
def test_info_table(self):
t = self.cwt_data.info_table
assert_equal(len(t.colnames), 7)
assert_equal(len(t), 13)
def test_sub(self):
h = self.cwt.history
diff = h[7] - h[5]
assert_equal(diff.support_3d.data.sum(), 0)
assert_allclose(diff.model.data.sum(), 20.4132906267)
def test_io(self, tmpdir):
filename = str(tmpdir / "test-cwt.fits")
self.cwt_data.write(filename=filename, overwrite=True)
approx = Map.read(filename, hdu="APPROX")
assert_allclose(approx.data[100, 100], self.cwt_data._approx[100, 100])
assert_allclose(approx.data[36, 63], self.cwt_data._approx[36, 63])
transform_2d = Map.read(filename, hdu="TRANSFORM_2D")
assert_allclose(transform_2d.data[100, 100],
self.cwt_data.transform_2d.data[100, 100])
assert_allclose(transform_2d.data[36, 63],
self.cwt_data.transform_2d.data[36, 63])
