def pca_common_script(
paths_in, dtype, sub, weights, correlation, reference, output_prefix,
num_levels, dump_pcs, unit_name, unit
):
"""Shared code by tr-pca and tr-pca-geom.
This function has only one purpose: bugs common to tr-pca and tr-pca-geom
have to be fixed only once.
"""
if num_levels < 2:
raise ValueError("num_levels must be at least 2.")
# call pca routine in tracks.api
mtr = MultiTracksReader(paths_in, dtype, sub=sub)
cm, overlap, cbs = pca_levels(mtr, num_levels, weights, correlation, reference)
# dump some stuff to disk
dump_track("%s.cov" % output_prefix, cm.cov.ravel())
dump_track("%s.evals" % output_prefix, cm.evals)
dump_track("%s.sigmas" % output_prefix, cm.sigmas)
for i in xrange(len(cm.evals)):
dump_track("%s.mode.%07i" % (output_prefix, i), cm.evecs[:,i])
# compute the cosine contents and optionally write the principal components to disk
mtr = MultiTracksReader(paths_in, dtype, sub=sub)
for data in mtr.iter_buffers():
data = data["data"]
for level in xrange(num_levels):
if level==0 and dump_pcs:
cbs[level].project_data(data, output_prefix)
else:
cbs[level].project_data(data)
dump_track("%s.ccs" % output_prefix, cm.ccs)
dump_track("%s.cosamp" % output_prefix, cm.dot_data_cos/cm.sqnorms_cos)
# Print some nice screen output with the most relevant results
print "Overlap between the covariance of the first and the second half of the trajectory:"
print " %.2f %%" % (overlap*100)
print
white = (" "*len(unit_name))
for level in xrange(num_levels):
sigmas, sigmas_err, ccs, ccs_err, overlap_multi = cbs[level].get_averages()
print "Level %i: averages over blocks with size total/%i=%i" % (
level, 2**level, mtr.shortest/(2**level)
)
if level > 0:
print "RMSD/RMSA of the covariance matrices of each block: %.2f %%" % (overlap_multi*100)
print
if level == 0:
print " Sigma [%s] Cosine content " % unit_name
print " ------------%s ---------------- " % white
for i in xrange(len(cm.sigmas)):
print " %7i %9.3e%s %8.2f %%" % (
i, sigmas[i]/unit, white, ccs[i]*100
)
else:
print " Sigma [%s] Cosine content " % unit_name
print " ----------------------%s ---------------------- " % white
for i in xrange(len(cm.sigmas)):
print " %7i %9.3e +- %9.3e%s %8.2f %% +- %6.2f %%" % (
i, sigmas[i]/unit, sigmas_err[i]/unit, white,
ccs[i]*100, ccs_err[i]*100
)
print
print
if level > 0:
dump_track("%s.ccs.%07i" % (output_prefix, level), ccs)
dump_track("%s.ccs_err.%07i" % (output_prefix, level), ccs_err)
dump_track("%s.sigmas.%07i" % (output_prefix, level), sigmas)
dump_track("%s.sigmas_err.%07i" % (output_prefix, level), sigmas_err)
return cm.mean
def dump_to_tracks(self, prefix):
dump_track("%s.x" % prefix, self.data[0])
dump_track("%s.y" % prefix, self.data[1])
dump_track("%s.z" % prefix, self.data[2])
def test_duration(self):
dump_track("test", numpy.arange(50))
self.assertEqual(parse_x_duration("test"), 49)
def test_length(self):
dump_track("test", numpy.arange(50))
self.assertEqual(parse_x_length("test"), 50)
def test_step(self):
dump_track("test", numpy.arange(50))
self.assertEqual(parse_x_step("test"), 1)
