# load btchroma, create mask
btchroma = sio.loadmat('/home/thierry/Columbia/covers80/coversongs/covers32kENmats/john_lennon+Double_Fantasy+05-I_m_Losing_You.mp3.mat')['btchroma']
p1=185;p2=p1+15;mask=np.ones(btchroma.shape);mask[:,p1:p2]=0.
# method: lintrans
recon_lintrans,tmp = IMPUTATION.lintransform_patch(btchroma,mask,p1,p2,win=1)
err_lintrans = evaluation.recon_error(btchroma,mask,recon_lintrans)
# method: random
recon_rand = IMPUTATION.random_patch(btchroma,mask,p1,p2)
err_rand = evaluation.recon_error(btchroma,mask,recon_rand)
# method: average
recon_avg = IMPUTATION.average_patch(btchroma,mask,p1,p2,win=1)
err_avg = evaluation.recon_error(btchroma,mask,recon_avg)
# method: NN
recon_nn,tmp = IMPUTATION.knn_patch(btchroma,mask,p1,p2,win=8,measure='eucl')
err_nn = evaluation.recon_error(btchroma,mask,recon_nn)
# method: SIPLCA
recon_siplca = IMPUTATION_PLCA.SIPLCA_mask.analyze((btchroma*mask).copy(),
50,mask,win=40,
convergence_thresh=1e-15)[4]
err_siplca = evaluation.recon_error(btchroma,mask,recon_siplca)
# PLOT
pargs = {'origin':'lower','interpolation':'nearest','cmap':P.cm.gray_r,
'aspect':'auto'}
pos1 = p1-10
pos2 = p2+70
P.figure()
P.hold(True)
# original and masked
recon_lintrans, tmp = IMPUTATION.lintransform_patch(btchroma,
mask,
p1,
p2,
win=1)
err_lintrans = evaluation.recon_error(btchroma, mask, recon_lintrans)
# method: random
recon_rand = IMPUTATION.random_patch(btchroma, mask, p1, p2)
err_rand = evaluation.recon_error(btchroma, mask, recon_rand)
# method: average
recon_avg = IMPUTATION.average_patch(btchroma, mask, p1, p2, win=1)
err_avg = evaluation.recon_error(btchroma, mask, recon_avg)
# method: NN
recon_nn, tmp = IMPUTATION.knn_patch(btchroma,
mask,
p1,
p2,
win=8,
measure='eucl')
err_nn = evaluation.recon_error(btchroma, mask, recon_nn)
# method: SIPLCA
recon_siplca = IMPUTATION_PLCA.SIPLCA_mask.analyze((btchroma * mask).copy(),
50,
mask,
win=40,
convergence_thresh=1e-15)[4]
err_siplca = evaluation.recon_error(btchroma, mask, recon_siplca)
# PLOT
pargs = {
'origin': 'lower',
'interpolation': 'nearest',
def plot_oneexample(btchroma,
mask,
p1,
p2,
methods=(),
methods_args=None,
measures=(),
subplot=None,
plotrange=None,
**kwargs):
"""
Utility function to plot the result of many methods on one example.
Methods and arguments are the same as when we test on a whole dataset.
First two plots are always original then original*mask.
INPUT
btchroma - btchroma
mask - binary mask, 0 = hidden
p1 - first masked column
p2 - last masekdcolumn + 1
methods - list of methods (lintrans,random,....)
bethods_args - list of dictionary containing arguments
measures - list of measures, put in title (we sort according to first)
subplot - if None, images in one column
plotrange - if we dont want to plot everything set to (pos1,pos2)
kwargs - extra arguments passed to plotall
"""
from plottools import plotall
# inits and sanity checks
for meas in measures:
assert meas in ('eucl', 'kl', 'cos', 'dent', 'lhalf', 'ddiff', 'jdiff',
'thresh', 'leven',
'condent'), 'unknown measure: ' + meas
nmethods = len(methods)
assert nmethods > 0, 'we need at least one method...!'
if methods_args is None:
methods_args = [{}] * len(methods)
if subplot is None:
subplot = (nmethods + 2, 1)
assert np.zeros(subplot).size == len(
methods) + 2, 'wrong subplot, forgot original and masked original?'
if plotrange is None:
plotrange = (0, btchroma.shape[1])
# impute
errs = []
recons = []
########## ALGORITHM DEPENDENT
for im, method in enumerate(methods):
if method == 'random':
recon = IMPUTATION.random_patch(btchroma, mask, p1, p2,
**methods_args[im])
elif method == 'randomfromsong':
recon = IMPUTATION.random_patch_from_song(btchroma, mask, p1, p2,
**methods_args[im])
elif method == 'average':
recon = IMPUTATION.average_patch(btchroma, mask, p1, p2,
**methods_args[im])
elif method == 'averageall':
recon = IMPUTATION.average_patch_all(btchroma, mask, p1, p2,
**methods_args[im])
elif method == 'codebook':
recon, used_codes = IMPUTATION.codebook_patch(
btchroma, mask, p1, p2, **methods_args[im])
elif method == 'knn_eucl':
recon, used_cols = IMPUTATION.knn_patch(btchroma,
mask,
p1,
p2,
measure='eucl',
**methods_args[im])
elif method == 'knn_kl':
recon, used_cols = IMPUTATION.knn_patch(btchroma,
mask,
p1,
p2,
measure='kl',
**methods_args[im])
elif method == 'lintrans':
recon, proj = IMPUTATION.lintransform_patch(
btchroma, mask, p1, p2, **methods_args[im])
elif method == 'kalman':
import kalman_toolbox as KALMAN
recon = KALMAN.imputation(btchroma, p1, p2, **methods_args[im])
elif method == 'hmm':
recon, recon2, hmm = IMPUTATION_HMM.imputation(
btchroma * mask, p1, p2, **methods_args[im])
elif method == 'siplca':
rank = methods_args[im]['rank']
res = IMPUTATION_PLCA.SIPLCA_mask.analyze((btchroma * mask).copy(),
rank,
mask,
convergence_thresh=1e-15,
**methods_args[im])
W, Z, H, norm, recon, logprob = res
elif method == 'siplca2':
rank = methods_args[im]['rank']
res = IMPUTATION_PLCA.SIPLCA2_mask.analyze(
(btchroma * mask).copy(),
rank,
mask,
convergence_thresh=1e-15,
**methods_args[im])
W, Z, H, norm, recon, logprob = res
else:
print 'unknown method:', method
return
# compute errors
recons.append(recon.copy())
errs.append(recon_error(btchroma, mask, recon, 'all'))
# all methods computed
# now we sort the results according to first measure
main_errs = map(lambda x: x[measures[0]], errs)
order = np.array(np.argsort(main_errs))
methods = map(lambda i: methods[i], order)
errs = map(lambda i: errs[i], order)
recons = map(lambda i: recons[i], order)
# plot
import pylab as P
P.figure()
pos1 = plotrange[0]
pos2 = plotrange[1]
ims = [btchroma[:, pos1:pos2], (btchroma * mask)[:, pos1:pos2]]
ims.extend(map(lambda r: r[:, pos1:pos2], recons))
titles = ['original', 'original masked']
for imethod, method in enumerate(methods):
t = method + ':'
for m in measures:
t += ' ' + m + '=' + nice_nums(errs[imethod][m])
titles.append(t)
plotall(ims, subplot=subplot, title=titles, cmap='gray_r', colorbar=False)
P.show()
# analysis
print 'recon shape:',recon.shape
div_eucl = evaluation.recon_error(btchroma,mask,recon,measure='eucl')
div_kl = evaluation.recon_error(btchroma,mask,recon,measure='kl')
print 'eucl div:',div_eucl
print 'kl div:',div_kl
# comparison
import imputation as IMPUTATION
print 'average, win=3'
recon = IMPUTATION.average_patch(btchroma,mask,p1,p2,win=3)
div_eucl = evaluation.recon_error(btchroma,mask,recon,measure='eucl')
div_kl = evaluation.recon_error(btchroma,mask,recon,measure='kl')
print 'eucl div:',div_eucl
print 'kl div:',div_kl
print 'kkn eucl:'
recon,used_cols = IMPUTATION.knn_patch(btchroma,mask,p1,p2,win=7,measure='eucl')
div_eucl = evaluation.recon_error(btchroma,mask,recon,measure='eucl')
div_kl = evaluation.recon_error(btchroma,mask,recon,measure='kl')
print 'eucl div:',div_eucl
print 'kl div:',div_kl
print 'linear transform'
recon,proj = IMPUTATION.lintransform_patch(btchroma,mask,p1,p2,win=2)
div_eucl = evaluation.recon_error(btchroma,mask,recon,measure='eucl')
div_kl = evaluation.recon_error(btchroma,mask,recon,measure='kl')
print 'eucl div:',div_eucl
print 'kl div:',div_kl
def test_maskedpatch_on_dataset(dataset,
method='random',
ncols=2,
win=1,
rank=4,
codebook=None,
nstates=1,
verbose=1,
**kwargs):
"""
Dataset is either dir, or list of files
General method to test a method on a whole dataset for one masked column
Methods are:
- random
- randomfromsong
- average
- averageall
- codebook
- knn_eucl
- knn_kl
- knn_eucl_delta
- lintrans
- kalman
- hmm
- siplca
- siplca2
Used arguments vary based on the method. For SIPLCA, we can use **kwargs
to set priors.
RETURN
- all errors, as a list of dict
"""
# get all matfiles
if type(dataset).__name__ == 'str':
matfiles = get_all_matfiles(dataset)
else:
matfiles = dataset
# init
total_cnt = 0
all_errs = []
# some specific inits
if codebook != None and not type(codebook) == type([]):
codebook = [p.reshape(12, codebook.shape[1] / 12) for p in codebook]
print 'codebook in ndarray format transformed to list'
# iterate
for matfile in matfiles:
btchroma = sio.loadmat(matfile)['btchroma']
if len(btchroma.shape) < 2:
continue
if btchroma.shape[1] < MINLENGTH or np.isnan(btchroma).any():
continue
mask, p1, p2 = MASKING.random_patch_mask(btchroma, ncols=ncols, win=25)
########## ALGORITHM DEPENDENT
if method == 'random':
recon = IMPUTATION.random_patch(btchroma, mask, p1, p2)
elif method == 'randomfromsong':
recon = IMPUTATION.random_patch_from_song(btchroma, mask, p1, p2)
elif method == 'average':
recon = IMPUTATION.average_patch(btchroma, mask, p1, p2, win=win)
elif method == 'averageall':
recon = IMPUTATION.average_patch_all(btchroma, mask, p1, p2)
elif method == 'codebook':
recon, used_codes = IMPUTATION.codebook_patch(
btchroma, mask, p1, p2, codebook)
elif method == 'knn_eucl':
recon, used_cols = IMPUTATION.knn_patch(btchroma,
mask,
p1,
p2,
win=win,
measure='eucl')
elif method == 'knn_kl':
recon, used_cols = IMPUTATION.knn_patch(btchroma,
mask,
p1,
p2,
win=win,
measure='kl')
elif method == 'knn_eucl_delta':
recon, used_cols = IMPUTATION.knn_patch_delta(btchroma,
mask,
p1,
p2,
win=win,
measure='eucl')
elif method == 'lintrans':
recon, proj = IMPUTATION.lintransform_patch(btchroma,
mask,
p1,
p2,
win=win)
elif method == 'kalman':
import kalman_toolbox as KALMAN
recon = KALMAN.imputation(btchroma,
p1,
p2,
dimstates=nstates,
**kwargs)
elif method == 'hmm':
recon, recon2, hmm = IMPUTATION_HMM.imputation(btchroma * mask,
p1,
p2,
nstates=nstates,
**kwargs)
elif method == 'siplca':
res = IMPUTATION_PLCA.SIPLCA_mask.analyze((btchroma * mask).copy(),
rank,
mask,
win=win,
convergence_thresh=1e-15,
**kwargs)
W, Z, H, norm, recon, logprob = res
elif method == 'siplca2':
res = IMPUTATION_PLCA.SIPLCA2_mask.analyze(
(btchroma * mask).copy(),
rank,
mask,
win=win,
convergence_thresh=1e-15,
**kwargs)
W, Z, H, norm, recon, logprob = res
else:
print 'unknown method:', method
return
########## ALGORITHM DEPENDENT END
# measure recon
errs = recon_error(btchroma, mask, recon, measure='all')
all_errs.append(errs)
total_cnt += 1
# done
if verbose > 0:
print 'number of songs tested:', total_cnt
for meas in np.sort(all_errs[0].keys()):
errs = map(lambda d: d[meas], all_errs)
print 'average', meas, '=', np.mean(errs), '(', np.std(errs), ')'
return all_errs
def plot_oneexample(btchroma,mask,p1,p2,methods=(),methods_args=None,
measures=(),subplot=None,plotrange=None,**kwargs):
"""
Utility function to plot the result of many methods on one example.
Methods and arguments are the same as when we test on a whole dataset.
First two plots are always original then original*mask.
INPUT
btchroma - btchroma
mask - binary mask, 0 = hidden
p1 - first masked column
p2 - last masekdcolumn + 1
methods - list of methods (lintrans,random,....)
bethods_args - list of dictionary containing arguments
measures - list of measures, put in title (we sort according to first)
subplot - if None, images in one column
plotrange - if we dont want to plot everything set to (pos1,pos2)
kwargs - extra arguments passed to plotall
"""
from plottools import plotall
# inits and sanity checks
for meas in measures:
assert meas in ('eucl','kl','cos','dent','lhalf','ddiff','jdiff',
'thresh','leven','condent'),'unknown measure: '+meas
nmethods = len(methods)
assert nmethods > 0,'we need at least one method...!'
if methods_args is None:
methods_args = [{}]*len(methods)
if subplot is None:
subplot = (nmethods+2,1)
assert np.zeros(subplot).size == len(methods)+2,'wrong subplot, forgot original and masked original?'
if plotrange is None:
plotrange = (0,btchroma.shape[1])
# impute
errs = []
recons = []
########## ALGORITHM DEPENDENT
for im,method in enumerate(methods):
if method == 'random':
recon = IMPUTATION.random_patch(btchroma,mask,p1,p2,**methods_args[im])
elif method == 'randomfromsong':
recon = IMPUTATION.random_patch_from_song(btchroma,mask,p1,p2,**methods_args[im])
elif method == 'average':
recon = IMPUTATION.average_patch(btchroma,mask,p1,p2,**methods_args[im])
elif method == 'averageall':
recon = IMPUTATION.average_patch_all(btchroma,mask,p1,p2,**methods_args[im])
elif method == 'codebook':
recon,used_codes = IMPUTATION.codebook_patch(btchroma,mask,p1,p2,**methods_args[im])
elif method == 'knn_eucl':
recon,used_cols = IMPUTATION.knn_patch(btchroma,mask,p1,p2,measure='eucl',**methods_args[im])
elif method == 'knn_kl':
recon,used_cols = IMPUTATION.knn_patch(btchroma,mask,p1,p2,measure='kl',**methods_args[im])
elif method == 'lintrans':
recon,proj = IMPUTATION.lintransform_patch(btchroma,mask,p1,p2,**methods_args[im])
elif method == 'kalman':
import kalman_toolbox as KALMAN
recon = KALMAN.imputation(btchroma,p1,p2,**methods_args[im])
elif method == 'hmm':
recon,recon2,hmm = IMPUTATION_HMM.imputation(btchroma*mask,p1,p2,
**methods_args[im])
elif method == 'siplca':
rank = methods_args[im]['rank']
res = IMPUTATION_PLCA.SIPLCA_mask.analyze((btchroma*mask).copy(),
rank,mask,
convergence_thresh=1e-15,
**methods_args[im])
W, Z, H, norm, recon, logprob = res
elif method == 'siplca2':
rank = methods_args[im]['rank']
res = IMPUTATION_PLCA.SIPLCA2_mask.analyze((btchroma*mask).copy(),
rank,mask,
convergence_thresh=1e-15,
**methods_args[im])
W, Z, H, norm, recon, logprob = res
else:
print 'unknown method:',method
return
# compute errors
recons.append( recon.copy() )
errs.append( recon_error(btchroma,mask,recon,'all') )
# all methods computed
# now we sort the results according to first measure
main_errs = map(lambda x: x[measures[0]], errs)
order = np.array(np.argsort(main_errs))
methods = map(lambda i: methods[i], order)
errs = map(lambda i: errs[i], order)
recons = map(lambda i: recons[i], order)
# plot
import pylab as P
P.figure()
pos1 = plotrange[0]
pos2 = plotrange[1]
ims = [btchroma[:,pos1:pos2],(btchroma*mask)[:,pos1:pos2]]
ims.extend( map(lambda r: r[:,pos1:pos2], recons) )
titles = ['original', 'original masked']
for imethod,method in enumerate(methods):
t = method + ':'
for m in measures:
t += ' ' + m + '=' + nice_nums(errs[imethod][m])
titles.append(t)
plotall(ims,subplot=subplot,title=titles,cmap='gray_r',colorbar=False)
P.show()
def test_maskedpatch_on_dataset(dataset,method='random',ncols=2,win=1,rank=4,codebook=None,nstates=1,verbose=1,**kwargs):
"""
Dataset is either dir, or list of files
General method to test a method on a whole dataset for one masked column
Methods are:
- random
- randomfromsong
- average
- averageall
- codebook
- knn_eucl
- knn_kl
- knn_eucl_delta
- lintrans
- kalman
- hmm
- siplca
- siplca2
Used arguments vary based on the method. For SIPLCA, we can use **kwargs
to set priors.
RETURN
- all errors, as a list of dict
"""
# get all matfiles
if type(dataset).__name__ == 'str':
matfiles = get_all_matfiles(dataset)
else:
matfiles = dataset
# init
total_cnt = 0
all_errs = []
# some specific inits
if codebook != None and not type(codebook) == type([]):
codebook = [p.reshape(12,codebook.shape[1]/12) for p in codebook]
print 'codebook in ndarray format transformed to list'
# iterate
for matfile in matfiles:
btchroma = sio.loadmat(matfile)['btchroma']
if len(btchroma.shape) < 2:
continue
if btchroma.shape[1] < MINLENGTH or np.isnan(btchroma).any():
continue
mask,p1,p2 = MASKING.random_patch_mask(btchroma,ncols=ncols,win=25)
########## ALGORITHM DEPENDENT
if method == 'random':
recon = IMPUTATION.random_patch(btchroma,mask,p1,p2)
elif method == 'randomfromsong':
recon = IMPUTATION.random_patch_from_song(btchroma,mask,p1,p2)
elif method == 'average':
recon = IMPUTATION.average_patch(btchroma,mask,p1,p2,win=win)
elif method == 'averageall':
recon = IMPUTATION.average_patch_all(btchroma,mask,p1,p2)
elif method == 'codebook':
recon,used_codes = IMPUTATION.codebook_patch(btchroma,mask,p1,p2,codebook)
elif method == 'knn_eucl':
recon,used_cols = IMPUTATION.knn_patch(btchroma,mask,p1,p2,win=win,measure='eucl')
elif method == 'knn_kl':
recon,used_cols = IMPUTATION.knn_patch(btchroma,mask,p1,p2,win=win,measure='kl')
elif method == 'knn_eucl_delta':
recon,used_cols = IMPUTATION.knn_patch_delta(btchroma,mask,p1,p2,win=win,measure='eucl')
elif method == 'lintrans':
recon,proj = IMPUTATION.lintransform_patch(btchroma,mask,p1,p2,win=win)
elif method == 'kalman':
import kalman_toolbox as KALMAN
recon = KALMAN.imputation(btchroma,p1,p2,
dimstates=nstates,**kwargs)
elif method == 'hmm':
recon,recon2,hmm = IMPUTATION_HMM.imputation(btchroma*mask,p1,p2,
nstates=nstates,
**kwargs)
elif method == 'siplca':
res = IMPUTATION_PLCA.SIPLCA_mask.analyze((btchroma*mask).copy(),
rank,mask,win=win,
convergence_thresh=1e-15,
**kwargs)
W, Z, H, norm, recon, logprob = res
elif method == 'siplca2':
res = IMPUTATION_PLCA.SIPLCA2_mask.analyze((btchroma*mask).copy(),
rank,mask,win=win,
convergence_thresh=1e-15,
**kwargs)
W, Z, H, norm, recon, logprob = res
else:
print 'unknown method:',method
return
########## ALGORITHM DEPENDENT END
# measure recon
errs = recon_error(btchroma,mask,recon,measure='all')
all_errs.append( errs )
total_cnt += 1
# done
if verbose > 0:
print 'number of songs tested:',total_cnt
for meas in np.sort(all_errs[0].keys()):
errs = map(lambda d: d[meas], all_errs)
print 'average',meas,'=',np.mean(errs),'(',np.std(errs),')'
return all_errs
s = 'method: linear transform\n'
s += 'eucl = ' + nice_nums(errs['eucl']) + '\n'
s += 'd1/2 = ' + nice_nums(errs['lhalf']) + '\n'
s += 'cond. ent. = ' + nice_nums(errs['condent']) + '\n'
s += 'delta diff. = ' + nice_nums(errs['ddiff']) + '\n'
s += 'D-ENT = ' + nice_nums(errs['dent']) + '\n'
s += 'Jensen diff. = ' + nice_nums(errs['jdiff']) + '\n'
originals.append( bt[:,p1:p2].copy() )
recons.append( recon[:,p1:p2].copy() )
texts.append(s)
# SMOOTH RECONSTRUCTION
bt = sio.loadmat('/home/thierry/Columbia/covers80/coversongs/covers32kENmats/metallica+Live_Shit_Binge_And_Purge_Disc_3_+09-Stone_Cold_Crazy.mp3.mat')['btchroma']
mask,p1,p2 = masking.random_patch_mask(bt,ncols=15,win=25)
recon,tmp = IMPUTATION.knn_patch(bt,mask,p1,p2,measure='eucl',win=8)
errs = evaluation.recon_error(bt,mask,recon)
s = 'method: NN\n'
s += 'eucl = ' + nice_nums(errs['eucl']) + '\n'
s += 'd1/2 = ' + nice_nums(errs['lhalf']) + '\n'
s += 'cond. ent. = ' + nice_nums(errs['condent']) + '\n'
s += 'delta diff. = ' + nice_nums(errs['ddiff']) + '\n'
s += 'D-ENT = ' + nice_nums(errs['dent']) + '\n'
s += 'Jensen diff. = ' + nice_nums(errs['jdiff']) + '\n'
originals.append( bt[:,p1:p2].copy() )
recons.append( recon[:,p1:p2].copy() )
texts.append(s)
# RANDOM RECON
bt = sio.loadmat('/home/thierry/Columbia/covers80/coversongs/covers32k/Caroline_No/beach_boys+Pet_Sounds+13-Caroline_No.mp3.mat')['btchroma']
div_eucl = evaluation.recon_error(btchroma, mask, recon, measure='eucl')
div_kl = evaluation.recon_error(btchroma, mask, recon, measure='kl')
print 'eucl div:', div_eucl
print 'kl div:', div_kl
# comparison
import imputation as IMPUTATION
print 'average, win=3'
recon = IMPUTATION.average_patch(btchroma, mask, p1, p2, win=3)
div_eucl = evaluation.recon_error(btchroma, mask, recon, measure='eucl')
div_kl = evaluation.recon_error(btchroma, mask, recon, measure='kl')
print 'eucl div:', div_eucl
print 'kl div:', div_kl
print 'kkn eucl:'
recon, used_cols = IMPUTATION.knn_patch(btchroma,
mask,
p1,
p2,
win=7,
measure='eucl')
div_eucl = evaluation.recon_error(btchroma, mask, recon, measure='eucl')
div_kl = evaluation.recon_error(btchroma, mask, recon, measure='kl')
print 'eucl div:', div_eucl
print 'kl div:', div_kl
print 'linear transform'
recon, proj = IMPUTATION.lintransform_patch(btchroma, mask, p1, p2, win=2)
div_eucl = evaluation.recon_error(btchroma, mask, recon, measure='eucl')
div_kl = evaluation.recon_error(btchroma, mask, recon, measure='kl')
print 'eucl div:', div_eucl
print 'kl div:', div_kl