def generateData(dsList, evals, CrE_A):
res = {}
D = set(i.dim for i in dsList).pop() # should have only one element
#if D == 3:
#set_trace()
bestalgentries = bestalg.loadBestAlgorithm(dsList.isBiobjective())
for fun, tmpdsList in dsList.dictByFunc().iteritems():
assert len(tmpdsList) == 1
entry = tmpdsList[0]
bestalgentry = bestalgentries[(D, fun)]
#ERT_A
f_A = detf(entry, evals)
ERT_best = detERT(bestalgentry, f_A)
ERT_A = detERT(entry, f_A)
nextbestf = []
for i in f_A:
if i == 0.:
nextbestf.append(0.)
else:
tmp = bestalgentry.target[bestalgentry.target < i]
try:
nextbestf.append(tmp[0])
except IndexError:
nextbestf.append(i * 10.**(-0.2)) # TODO: this is a hack
ERT_best_nextbestf = detERT(bestalgentry, nextbestf)
for i in range(len(ERT_A)):
# nextbestf[i] >= f_thresh: this is tested because if it is not true
# ERT_best_nextbestf[i] is supposed to be infinite.
if nextbestf[i] >= f_thresh and ERT_best_nextbestf[i] < evals[i]: # is different from the specification...
ERT_A[i] = evals[i]
# For test purpose:
#if fun % 10 == 0:
# ERT_A[-2] = 1.
# ERT_best[-2] = np.inf
ERT_A = np.array(ERT_A)
ERT_best = np.array(ERT_best)
loss_A = np.exp(CrE_A) * ERT_A / ERT_best
assert (np.isnan(loss_A) == False).all()
#set_trace()
#if np.isnan(loss_A).any() or np.isinf(loss_A).any() or (loss_A == 0.).any():
# txt = 'Problem with entry %s' % str(entry)
# warnings.warn(txt)
# #set_trace()
res[fun] = loss_A
return res
def main(dictAlg, isBiobjective, order=None, outputdir='.', info='default',
dimension=None, verbose=True):
"""Generates a figure showing the performance of algorithms.
From a dictionary of :py:class:`DataSetList` sorted by algorithms,
generates the cumulative distribution function of the bootstrap
distribution of ERT for algorithms on multiple functions for
multiple targets altogether.
:param dict dictAlg: dictionary of :py:class:`DataSetList` instances
one instance is equivalent to one algorithm,
:param list targets: target function values
:param list order: sorted list of keys to dictAlg for plotting order
:param str outputdir: output directory
:param str info: output file name suffix
:param bool verbose: controls verbosity
"""
global x_limit # late assignment of default, because it can be set to None in config
global divide_by_dimension # not fully implemented/tested yet
if 'x_limit' not in globals() or x_limit is None:
x_limit = x_limit_default
tmp = pp.dictAlgByDim(dictAlg)
# tmp = pp.DictAlg(dictAlg).by_dim()
if len(tmp) != 1 and dimension is None:
raise ValueError('We never integrate over dimension.')
if dimension is not None:
if dimension not in tmp.keys():
raise ValueError('dimension %d not in dictAlg dimensions %s'
% (dimension, str(tmp.keys())))
tmp = {dimension: tmp[dimension]}
dim = tmp.keys()[0]
divisor = dim if divide_by_dimension else 1
algorithms_with_data = [a for a in dictAlg.keys() if dictAlg[a] != []]
dictFunc = pp.dictAlgByFun(dictAlg)
# Collect data
# Crafting effort correction: should we consider any?
CrEperAlg = {}
for alg in algorithms_with_data:
CrE = 0.
if 1 < 3 and dictAlg[alg][0].algId == 'GLOBAL':
tmp = dictAlg[alg].dictByNoise()
assert len(tmp.keys()) == 1
if tmp.keys()[0] == 'noiselessall':
CrE = 0.5117
elif tmp.keys()[0] == 'nzall':
CrE = 0.6572
CrEperAlg[alg] = CrE
if CrE != 0.0:
print 'Crafting effort for', alg, 'is', CrE
dictData = {} # list of (ert per function) per algorithm
dictMaxEvals = {} # list of (maxevals per function) per algorithm
bestERT = [] # best ert per function
# funcsolved = [set()] * len(targets) # number of functions solved per target
xbest2009 = []
maxevalsbest2009 = []
for f, dictAlgperFunc in dictFunc.iteritems():
if function_IDs and f not in function_IDs:
continue
# print target_values((f, dim))
for j, t in enumerate(target_values((f, dim))):
# for j, t in enumerate(genericsettings.current_testbed.ecdf_target_values(1e2, f)):
# funcsolved[j].add(f)
for alg in algorithms_with_data:
x = [np.inf] * perfprofsamplesize
runlengthunsucc = []
try:
entry = dictAlgperFunc[alg][0] # one element per fun and per dim.
evals = entry.detEvals([t])[0]
assert entry.dim == dim
runlengthsucc = evals[np.isnan(evals) == False] / divisor
runlengthunsucc = entry.maxevals[np.isnan(evals)] / divisor
if len(runlengthsucc) > 0:
x = toolsstats.drawSP(runlengthsucc, runlengthunsucc,
percentiles=[50],
samplesize=perfprofsamplesize)[1]
except (KeyError, IndexError):
#set_trace()
warntxt = ('Data for algorithm %s on function %d in %d-D '
% (alg, f, dim)
+ 'are missing.\n')
warnings.warn(warntxt)
dictData.setdefault(alg, []).extend(x)
dictMaxEvals.setdefault(alg, []).extend(runlengthunsucc)
displaybest2009 = not isBiobjective #disabled until we find the bug
if displaybest2009:
#set_trace()
bestalgentries = bestalg.loadBestAlgorithm(isBiobjective)
bestalgentry = bestalgentries[(dim, f)]
bestalgevals = bestalgentry.detEvals(target_values((f, dim)))
# print bestalgevals
for j in range(len(bestalgevals[0])):
if bestalgevals[1][j]:
evals = bestalgevals[0][j]
#set_trace()
assert dim == bestalgentry.dim
runlengthsucc = evals[np.isnan(evals) == False] / divisor
runlengthunsucc = bestalgentry.maxevals[bestalgevals[1][j]][np.isnan(evals)] / divisor
x = toolsstats.drawSP(runlengthsucc, runlengthunsucc,
percentiles=[50],
samplesize=perfprofsamplesize)[1]
else:
x = perfprofsamplesize * [np.inf]
runlengthunsucc = []
xbest2009.extend(x)
maxevalsbest2009.extend(runlengthunsucc)
if order is None:
order = dictData.keys()
# Display data
lines = []
if displaybest2009:
args = {'ls': '-', 'linewidth': 6, 'marker': 'D', 'markersize': 11.,
'markeredgewidth': 1.5, 'markerfacecolor': refcolor,
'markeredgecolor': refcolor, 'color': refcolor,
'label': 'best 2009', 'zorder': -1}
lines.append(plotdata(np.array(xbest2009), x_limit, maxevalsbest2009,
CrE = 0., **args))
def algname_to_label(algname, dirname=None):
"""to be extended to become generally useful"""
if isinstance(algname, (tuple, list)): # not sure this is needed
return ' '.join([str(name) for name in algname])
return str(algname)
for i, alg in enumerate(order):
try:
data = dictData[alg]
maxevals = dictMaxEvals[alg]
except KeyError:
continue
args = styles[(i) % len(styles)]
args['linewidth'] = 1.5
args['markersize'] = 12.
args['markeredgewidth'] = 1.5
args['markerfacecolor'] = 'None'
args['markeredgecolor'] = args['color']
args['label'] = algname_to_label(alg)
#args['markevery'] = perfprofsamplesize # option available in latest version of matplotlib
#elif len(show_algorithms) > 0:
#args['color'] = 'wheat'
#args['ls'] = '-'
#args['zorder'] = -1
# plotdata calls pprldistr.plotECDF which calls ppfig.plotUnifLog... which does the work
lines.append(plotdata(np.array(data), x_limit, maxevals,
CrE=CrEperAlg[alg], **args))
labels, handles = plotLegend(lines, x_limit)
if True: # isLateXLeg:
fileName = os.path.join(outputdir,'pprldmany_%s.tex' % (info))
with open(fileName, 'w') as f:
f.write(r'\providecommand{\nperfprof}{7}')
algtocommand = {} # latex commands
for i, alg in enumerate(order):
tmp = r'\alg%sperfprof' % pptex.numtotext(i)
f.write(r'\providecommand{%s}{\StrLeft{%s}{\nperfprof}}' %
(tmp, toolsdivers.str_to_latex(
toolsdivers.strip_pathname2(algname_to_label(alg)))))
algtocommand[algname_to_label(alg)] = tmp
if displaybest2009:
tmp = r'\algzeroperfprof'
f.write(r'\providecommand{%s}{best 2009}' % (tmp))
algtocommand['best 2009'] = tmp
commandnames = []
for label in labels:
commandnames.append(algtocommand[label])
# f.write(headleg)
if len(order) > 28: # latex sidepanel won't work well for more than 25 algorithms, but original labels are also clipped
f.write(r'\providecommand{\perfprofsidepanel}{\mbox{%s}\vfill\mbox{%s}}'
% (commandnames[0], commandnames[-1]))
else:
fontsize_command = r'\tiny{}' if len(order) > 19 else ''
f.write(r'\providecommand{\perfprofsidepanel}{{%s\mbox{%s}' %
(fontsize_command, commandnames[0])) # TODO: check len(labels) > 0
for i in range(1, len(labels)):
f.write('\n' + r'\vfill \mbox{%s}' % commandnames[i])
f.write('}}\n')
# f.write(footleg)
if verbose:
print 'Wrote right-hand legend in %s' % fileName
figureName = os.path.join(outputdir,'pprldmany_%s' % (info))
#beautify(figureName, funcsolved, x_limit*x_annote_factor, False, fileFormat=figformat)
beautify()
text = ppfig.consecutiveNumbers(sorted(dictFunc.keys()), 'f')
text += ',%d-D' % dim # TODO: this is strange when different dimensions are plotted
plt.text(0.01, 0.98, text, horizontalalignment="left",
verticalalignment="top", transform=plt.gca().transAxes)
if len(dictFunc) == 1:
plt.title(' '.join((str(dictFunc.keys()[0]),
genericsettings.current_testbed.short_names[dictFunc.keys()[0]])))
a = plt.gca()
plt.xlim(xmin=1e-0, xmax=x_limit**annotation_space_end_relative)
xticks, labels = plt.xticks()
tmp = []
for i in xticks:
tmp.append('%d' % round(np.log10(i)))
a.set_xticklabels(tmp)
if save_figure:
ppfig.saveFigure(figureName, verbose=verbose)
if len(dictFunc) == 1:
ppfig.save_single_functions_html(
os.path.join(outputdir, 'pprldmany'),
'', # algorithms names are clearly visible in the figure
add_to_names='_%02dD' %(dim),
algorithmCount=ppfig.AlgorithmCount.NON_SPECIFIED
)
if close_figure:
plt.close()
def main(dictAlg, sortedAlgs, isBiobjective, outputdir='.', verbose=True, function_targets_line=True): # [1, 13, 101]
"""Generate one table per func with results of multiple algorithms."""
"""Difference with the first version:
* numbers aligned using the decimal separator
* premices for dispersion measure
* significance test against best algorithm
* table width...
Takes ``targetsOfInterest`` from this file as "input argument" to compute
the desired target values. ``targetsOfInterest`` might be configured via
config.
"""
# TODO: method is long, terrible to read, split if possible
bestalgentries = bestalg.loadBestAlgorithm(isBiobjective)
# Sort data per dimension and function
dictData = {}
dsListperAlg = list(dictAlg[i] for i in sortedAlgs)
for n, entries in enumerate(dsListperAlg):
tmpdictdim = entries.dictByDim()
for d in tmpdictdim:
tmpdictfun = tmpdictdim[d].dictByFunc()
for f in tmpdictfun:
dictData.setdefault((d, f), {})[n] = tmpdictfun[f]
nbtests = len(dictData)
funInfos = ppfigparam.read_fun_infos(isBiobjective)
for df in dictData:
# Generate one table per df
# first update targets for each dimension-function pair if needed:
targets = targetsOfInterest((df[1], df[0]))
targetf = targets[-1]
# best 2009
refalgentry = bestalgentries[df]
refalgert = refalgentry.detERT(targets)
refalgevals = (refalgentry.detEvals((targetf, ))[0][0])
refalgnbruns = len(refalgevals)
refalgnbsucc = numpy.sum(numpy.isnan(refalgevals) == False)
# Process the data
# The following variables will be lists of elements each corresponding
# to an algorithm
algnames = []
#algdata = []
algerts = []
algevals = []
algdisp = []
algnbsucc = []
algnbruns = []
algmedmaxevals = []
algmedfinalfunvals = []
algtestres = []
algentries = []
for n in sorted(dictData[df].keys()):
entries = dictData[df][n]
# the number of datasets for a given dimension and function (df)
# should be strictly 1. TODO: find a way to warn
# TODO: do this checking before... why wasn't it triggered by ppperprof?
if len(entries) > 1:
print entries
txt = ("There is more than a single entry associated with "
"folder %s on %d-D f%d." % (sortedAlgs[n], df[0], df[1]))
raise Exception(txt)
entry = entries[0]
algentries.append(entry)
algnames.append(sortedAlgs[n])
evals = entry.detEvals(targets)
#tmpdata = []
tmpdisp = []
tmpert = []
for i, e in enumerate(evals):
succ = (numpy.isnan(e) == False)
ec = e.copy() # note: here was the previous bug (changes made in e also appeared in evals !)
ec[succ == False] = entry.maxevals[succ == False]
ert = toolsstats.sp(ec, issuccessful=succ)[0]
#tmpdata.append(ert/refalgert[i])
if succ.any():
tmp = toolsstats.drawSP(ec[succ], entry.maxevals[succ == False],
[10, 50, 90], samplesize=samplesize)[0]
tmpdisp.append((tmp[-1] - tmp[0])/2.)
else:
tmpdisp.append(numpy.nan)
tmpert.append(ert)
algerts.append(tmpert)
algevals.append(evals)
#algdata.append(tmpdata)
algdisp.append(tmpdisp)
algmedmaxevals.append(numpy.median(entry.maxevals))
algmedfinalfunvals.append(numpy.median(entry.finalfunvals))
#algmedmaxevals.append(numpy.median(entry.maxevals)/df[0])
#algmedfinalfunvals.append(numpy.median(entry.finalfunvals))
algtestres.append(significancetest(refalgentry, entry, targets))
# determine success probability for Df = 1e-8
e = entry.detEvals((targetf ,))[0]
algnbsucc.append(numpy.sum(numpy.isnan(e) == False))
algnbruns.append(len(e))
# Process over all data
# find best values...
nalgs = len(dictData[df])
maxRank = 1 + numpy.floor(0.14 * nalgs) # number of algs to be displayed in bold
isBoldArray = [] # Point out the best values
algfinaldata = [] # Store median function values/median number of function evaluations
tmptop = getTopIndicesOfColumns(algerts, maxRank=maxRank)
for i, erts in enumerate(algerts):
tmp = []
for j, ert in enumerate(erts): # algi targetj
tmp.append(i in tmptop[j] or (nalgs > 7 and algerts[i][j] <= 3. * refalgert[j]))
isBoldArray.append(tmp)
algfinaldata.append((algmedfinalfunvals[i], algmedmaxevals[i]))
# significance test of best given algorithm against all others
best_alg_idx = numpy.array(algerts).argsort(0)[0, :] # indexed by target index
significance_versus_others = significance_all_best_vs_other(algentries, targets, best_alg_idx)[0]
# Create the table
table = []
tableHtml = []
spec = r'@{}c@{}|*{%d}{@{\,}r@{}X@{\,}}|@{}r@{}@{}l@{}' % (len(targets)) # in case StrLeft not working: replaced c@{} with l@{ }
spec = r'@{}c@{}|*{%d}{@{}r@{}X@{}}|@{}r@{}@{}l@{}' % (len(targets)) # in case StrLeft not working: replaced c@{} with l@{ }
extraeol = []
# Generate header lines
if with_table_heading:
header = funInfos[df[1]] if df[1] in funInfos.keys() else 'f%d' % df[1]
table.append([r'\multicolumn{%d}{@{\,}c@{\,}}{{\textbf{%s}}}'
% (2 * len(targets) + 2, header)])
extraeol.append('')
if function_targets_line is True or (function_targets_line and df[1] in function_targets_line):
if isinstance(targetsOfInterest, pproc.RunlengthBasedTargetValues):
curline = [r'\#FEs/D']
curlineHtml = ['<thead>\n<tr>\n<th>#FEs/D<br>REPLACEH</th>\n']
counter = 1
for i in targetsOfInterest.labels():
curline.append(r'\multicolumn{2}{@{}c@{}}{%s}' % i)
curlineHtml.append('<td>%s<br>REPLACE%d</td>\n' % (i, counter))
counter += 1
else:
curline = [r'$\Delta f_\mathrm{opt}$']
curlineHtml = ['<thead>\n<tr>\n<th>Δ f<sub>opt</sub><br>REPLACEH</th>\n']
counter = 1
for t in targets:
curline.append(r'\multicolumn{2}{@{\,}X@{\,}}{%s}'
% writeFEvals2(t, precision=1, isscientific=True))
curlineHtml.append('<td>%s<br>REPLACE%d</td>\n' % (writeFEvals2(t, precision=1, isscientific=True), counter))
counter += 1
# curline.append(r'\multicolumn{2}{@{\,}X@{}|}{%s}'
# % writeFEvals2(targets[-1], precision=1, isscientific=True))
curline.append(r'\multicolumn{2}{@{}l@{}}{\#succ}')
curlineHtml.append('<td>#succ<br>REPLACEF</td>\n</tr>\n</thead>\n')
table.append(curline)
extraeol.append(r'\hline')
# extraeol.append(r'\hline\arrayrulecolor{tableShade}')
curline = [r'ERT$_{\text{best}}$'] if with_table_heading else [r'\textbf{f%d}' % df[1]]
replaceValue = 'ERT<sub>best</sub>' if with_table_heading else ('<b>f%d</b>' % df[1])
curlineHtml = [item.replace('REPLACEH', replaceValue) for item in curlineHtml]
if isinstance(targetsOfInterest, pproc.RunlengthBasedTargetValues):
# write ftarget:fevals
counter = 1
for i in xrange(len(refalgert[:-1])):
temp="%.1e" %targetsOfInterest((df[1], df[0]))[i]
if temp[-2]=="0":
temp=temp[:-2]+temp[-1]
curline.append(r'\multicolumn{2}{@{}c@{}}{\textit{%s}:%s \quad}'
% (temp, writeFEvalsMaxPrec(refalgert[i], 2)))
replaceValue = '<i>%s</i>:%s' % (temp, writeFEvalsMaxPrec(refalgert[i], 2))
curlineHtml = [item.replace('REPLACE%d' % counter, replaceValue) for item in curlineHtml]
counter += 1
temp="%.1e" %targetsOfInterest((df[1], df[0]))[-1]
if temp[-2]=="0":
temp=temp[:-2]+temp[-1]
curline.append(r'\multicolumn{2}{@{}c@{}|}{\textit{%s}:%s }'
% (temp ,writeFEvalsMaxPrec(refalgert[-1], 2)))
replaceValue = '<i>%s</i>:%s' % (temp, writeFEvalsMaxPrec(refalgert[-1], 2))
curlineHtml = [item.replace('REPLACE%d' % counter, replaceValue) for item in curlineHtml]
else:
# write #fevals of the reference alg
counter = 1
for i in refalgert[:-1]:
curline.append(r'\multicolumn{2}{@{}c@{}}{%s \quad}'
% writeFEvalsMaxPrec(i, 2))
curlineHtml = [item.replace('REPLACE%d' % counter, writeFEvalsMaxPrec(i, 2)) for item in curlineHtml]
counter += 1
curline.append(r'\multicolumn{2}{@{}c@{}|}{%s}'
% writeFEvalsMaxPrec(refalgert[-1], 2))
curlineHtml = [item.replace('REPLACE%d' % counter, writeFEvalsMaxPrec(refalgert[-1], 2)) for item in curlineHtml]
# write the success ratio for the reference alg
tmp2 = numpy.sum(numpy.isnan(refalgevals) == False) # count the nb of success
curline.append('%d' % (tmp2))
if tmp2 > 0:
curline.append('/%d' % len(refalgevals))
replaceValue = '%d/%d' % (tmp2, len(refalgevals))
else:
replaceValue = '%d' % tmp2
curlineHtml = [item.replace('REPLACEF', replaceValue) for item in curlineHtml]
table.append(curline[:])
tableHtml.extend(curlineHtml[:])
tableHtml.append('<tbody>\n')
extraeol.append('')
#for i, gna in enumerate(zip((1, 2, 3), ('bla', 'blo', 'bli'))):
#print i, gna, gno
#set_trace()
# Format data
#if df == (5, 17):
#set_trace()
header = r'\providecommand{\ntables}{7}'
for i, alg in enumerate(algnames):
tableHtml.append('<tr>\n')
#algname, entries, irs, line, line2, succ, runs, testres1alg in zip(algnames,
#data, dispersion, isBoldArray, isItalArray, nbsucc, nbruns, testres):
commandname = r'\alg%stables' % numtotext(i)
# header += r'\providecommand{%s}{{%s}{}}' % (commandname, str_to_latex(strip_pathname(alg)))
header += r'\providecommand{%s}{\StrLeft{%s}{\ntables}}' % (commandname, str_to_latex(strip_pathname1(alg)))
curline = [commandname + r'\hspace*{\fill}'] # each list element becomes a &-separated table entry?
curlineHtml = ['<th>%s</th>\n' % str_to_latex(strip_pathname1(alg))]
for j, tmp in enumerate(zip(algerts[i], algdisp[i], # j is target index
isBoldArray[i], algtestres[i])):
ert, dispersion, isBold, testres = tmp
alignment = '@{\,}X@{\,}'
if j == len(algerts[i]) - 1:
alignment = '@{\,}X@{\,}|'
data = ert/refalgert[j]
# write star for significance against all other algorithms
str_significance_subsup = ''
str_significance_subsup_html = ''
if (len(best_alg_idx) > 0 and len(significance_versus_others) > 0 and
i == best_alg_idx[j] and nbtests * significance_versus_others[j][1] < 0.05):
logp = -numpy.ceil(numpy.log10(nbtests * significance_versus_others[j][1]))
logp = numpy.min((9, logp)) # not messing up the format and handling inf
str_significance_subsup = r"^{%s%s}" % (significance_vs_others_symbol, str(int(logp)) if logp > 1 else '')
str_significance_subsup_html = '<sup>%s%s</sup>' % (significance_vs_others_symbol_html, str(int(logp)) if logp > 1 else '')
# moved out of the above else: this was a bug!?
z, p = testres
if (nbtests * p) < 0.05 and data < 1. and z < 0.:
if not numpy.isinf(refalgert[j]):
tmpevals = algevals[i][j].copy()
tmpevals[numpy.isnan(tmpevals)] = algentries[i].maxevals[numpy.isnan(tmpevals)]
bestevals = refalgentry.detEvals(targets)
bestevals, bestalgalg = (bestevals[0][0], bestevals[1][0])
bestevals[numpy.isnan(bestevals)] = refalgentry.maxevals[bestalgalg][numpy.isnan(bestevals)]
tmpevals = numpy.array(sorted(tmpevals))[0:min(len(tmpevals), len(bestevals))]
bestevals = numpy.array(sorted(bestevals))[0:min(len(tmpevals), len(bestevals))]
#The conditions are now that ERT < ERT_best and
# all(sorted(FEvals_best) > sorted(FEvals_current)).
if numpy.isinf(refalgert[j]) or all(tmpevals < bestevals):
nbstars = -numpy.ceil(numpy.log10(nbtests * p))
# tmp2[-1] += r'$^{%s}$' % superscript
str_significance_subsup += r'_{%s%s}' % (significance_vs_ref_symbol,
str(int(nbstars)) if nbstars > 1 else '')
str_significance_subsup_html = '<sub>%s%s</sub>' % (significance_vs_ref_symbol_html,
str(int(nbstars)) if nbstars > 1 else '')
if str_significance_subsup:
str_significance_subsup = '$%s$' % str_significance_subsup
# format number in variable data
if numpy.isnan(data):
curline.append(r'\multicolumn{2}{%s}{.}' % alignment)
else:
if numpy.isinf(refalgert[j]):
curline.append(r'\multicolumn{2}{%s}{\textbf{%s}\mbox{\tiny (%s)}%s}'
% (alignment,
writeFEvalsMaxPrec(algerts[i][j], 2),
writeFEvalsMaxPrec(dispersion, precdispersion),
str_significance_subsup))
curlineHtml.append('<td sorttable_customkey=\"%f\"><b>%s</b> (%s)%s</td>\n'
% (algerts[i][j],
writeFEvalsMaxPrec(algerts[i][j], 2),
writeFEvalsMaxPrec(dispersion, precdispersion),
str_significance_subsup_html))
continue
tmp = writeFEvalsMaxPrec(data, precfloat, maxfloatrepr=maxfloatrepr)
tmpHtml = writeFEvalsMaxPrec(data, precfloat, maxfloatrepr=maxfloatrepr)
sortKey = data
if data >= maxfloatrepr or data < 0.01: # either inf or scientific notation
if numpy.isinf(data) and j == len(algerts[i]) - 1:
tmp += r'\,\textit{%s}' % writeFEvalsMaxPrec(algfinaldata[i][1], 0, maxfloatrepr=maxfloatrepr)
tmpHtml += '<i>%s</i>' % writeFEvalsMaxPrec(algfinaldata[i][1], 0, maxfloatrepr=maxfloatrepr)
sortKey = algfinaldata[i][1]
else:
tmp = writeFEvalsMaxPrec(data, precscien, maxfloatrepr=data)
if isBold:
tmpHtml = '<b>%s</b>' % tmp
tmp = r'\textbf{%s}' % tmp
if not numpy.isnan(dispersion):
tmpdisp = dispersion/refalgert[j]
if tmpdisp >= maxfloatrepr or tmpdisp < 0.005: # TODO: hack
tmpdisp = writeFEvalsMaxPrec(tmpdisp, precdispersion, maxfloatrepr=tmpdisp)
else:
tmpdisp = writeFEvalsMaxPrec(tmpdisp, precdispersion, maxfloatrepr=maxfloatrepr)
tmp += r'\mbox{\tiny (%s)}' % tmpdisp
tmpHtml += ' (%s)' % tmpdisp
curline.append(r'\multicolumn{2}{%s}{%s%s}' % (alignment, tmp, str_significance_subsup))
tmpHtml = tmpHtml.replace('$\infty$', '∞')
if (numpy.isinf(sortKey)):
sortKey = sys.maxint
curlineHtml.append('<td sorttable_customkey=\"%f\">%s%s</td>' % (sortKey, tmpHtml, str_significance_subsup_html))
else:
tmp2 = tmp.split('.', 1)
if len(tmp2) < 2:
tmp2.append('')
else:
tmp2[-1] = '.' + tmp2[-1]
if isBold:
tmp3 = []
tmp3html = []
for k in tmp2:
tmp3.append(r'\textbf{%s}' % k)
tmp3html.append('<b>%s</b>' % k)
tmp2 = tmp3
tmp2html = tmp3html
else:
tmp2html = []
tmp2html.extend(tmp2)
if not numpy.isnan(dispersion):
tmpdisp = dispersion/refalgert[j]
if tmpdisp >= maxfloatrepr or tmpdisp < 0.01:
tmpdisp = writeFEvalsMaxPrec(tmpdisp, precdispersion, maxfloatrepr=tmpdisp)
else:
tmpdisp = writeFEvalsMaxPrec(tmpdisp, precdispersion, maxfloatrepr=maxfloatrepr)
tmp2[-1] += (r'\mbox{\tiny (%s)}' % (tmpdisp))
tmp2html[-1] += ' (%s)' % tmpdisp
tmp2[-1] += str_significance_subsup
tmp2html[-1] += str_significance_subsup_html
curline.extend(tmp2)
tmp2html = ("").join(str(item) for item in tmp2html)
tmp2html = tmp2html.replace('$\infty$', '∞')
curlineHtml.append('<td sorttable_customkey=\"%f\">%s</td>' % (data, tmp2html))
curline.append('%d' % algnbsucc[i])
curline.append('/%d' % algnbruns[i])
table.append(curline)
curlineHtml.append('<td sorttable_customkey=\"%d\">%d/%d</td>\n' % (algnbsucc[i], algnbsucc[i], algnbruns[i]))
tableHtml.extend(curlineHtml[:])
extraeol.append('')
# Write table
res = tableXLaTeX(table, spec=spec, extraeol=extraeol)
try:
filename = os.path.join(outputdir, 'pptables_f%03d_%02dD.tex' % (df[1], df[0]))
f = open(filename, 'w')
f.write(header + '\n')
f.write(res)
res = ("").join(str(item) for item in tableHtml)
res = '\n<table class=\"sortable\" style=\"width:800px \">\n%s</table>\n<p/>\n' % res
if df[0] in (5, 20):
filename = os.path.join(outputdir, genericsettings.many_algorithm_file_name + '.html')
lines = []
with open(filename) as infile:
for line in infile:
if '<!--' + 'pptablesf%03d%02dDHtml' % (df[1], df[0]) + '-->' in line:
lines.append(res)
lines.append(line)
with open(filename, 'w') as outfile:
for line in lines:
outfile.write(line)
if verbose:
print 'Wrote table in %s' % filename
except:
raise
else:
f.close()
def main(dsList0, dsList1, dimsOfInterest, outputdir, info='', verbose=True):
"""One table per dimension, modified to fit in 1 page per table."""
#TODO: method is long, split if possible
dictDim0 = dsList0.dictByDim()
dictDim1 = dsList1.dictByDim()
alg0 = set(i[0] for i in dsList0.dictByAlg().keys()).pop().replace(genericsettings.extraction_folder_prefix, '')[0:3]
alg1 = set(i[0] for i in dsList1.dictByAlg().keys()).pop().replace(genericsettings.extraction_folder_prefix, '')[0:3]
open(os.path.join(outputdir, 'bbob_pproc_commands.tex'), 'a'
).write(r'\providecommand{\algorithmAshort}{%s}' % writeLabels(alg0) + '\n' +
r'\providecommand{\algorithmBshort}{%s}' % writeLabels(alg1) + '\n')
if info:
info = '_' + info
dims = set.intersection(set(dictDim0.keys()), set(dictDim1.keys()))
bestalgentries = bestalg.loadBestAlgorithm(dsList0.isBiobjective())
header = []
if isinstance(targetsOfInterest, pproc.RunlengthBasedTargetValues):
header = [r'\#FEs/D']
headerHtml = ['<thead>\n<tr>\n<th>#FEs/D</th>\n']
for label in targetsOfInterest.labels():
header.append(r'\multicolumn{2}{@{}c@{}}{%s}' % label)
headerHtml.append('<td>%s</td>\n' % label)
else:
header = [r'$\Delta f_\mathrm{opt}$']
headerHtml = ['<thead>\n<tr>\n<th>Δ f</th>\n']
for label in targetsOfInterest.labels():
header.append(r'\multicolumn{2}{@{\,}c@{\,}}{%s}' % label)
headerHtml.append('<td>%s</td>\n' % label)
header.append(r'\multicolumn{2}{@{}l@{}}{\#succ}')
headerHtml.append('<td>#succ</td>\n</tr>\n</thead>\n')
for d in dimsOfInterest: # TODO set as input arguments
table = [header]
tableHtml = headerHtml
extraeol = [r'\hline']
try:
dictFunc0 = dictDim0[d].dictByFunc()
dictFunc1 = dictDim1[d].dictByFunc()
except KeyError:
continue
funcs = set.union(set(dictFunc0.keys()), set(dictFunc1.keys()))
nbtests = len(funcs) * 2. #len(dimsOfInterest)
tableHtml.append('<tbody>\n')
for f in sorted(funcs):
tableHtml.append('<tr>\n')
targets = targetsOfInterest((f, d))
targetf = targets[-1]
bestalgentry = bestalgentries[(d, f)]
curline = [r'${\bf f_{%d}}$' % f]
curlineHtml = ['<th><b>f<sub>%d</sub></b></th>\n' % f]
bestalgdata = bestalgentry.detERT(targets)
bestalgevals, bestalgalgs = bestalgentry.detEvals(targets)
if isinstance(targetsOfInterest, pproc.RunlengthBasedTargetValues):
# write ftarget:fevals
for i in xrange(len(bestalgdata[:-1])):
temp = "%.1e" % targetsOfInterest((f, d))[i]
if temp[-2]=="0":
temp = temp[:-2]+temp[-1]
curline.append(r'\multicolumn{2}{@{}c@{}}{\textit{%s}:%s \quad}'
% (temp,writeFEvalsMaxPrec(bestalgdata[i], 2)))
curlineHtml.append('<td><i>%s</i>:%s</td>\n'
% (temp, writeFEvalsMaxPrec(bestalgdata[i], 2)))
temp = "%.1e" % targetsOfInterest((f, d))[-1]
if temp[-2]=="0":
temp = temp[:-2]+temp[-1]
curline.append(r'\multicolumn{2}{@{}c@{}|}{\textit{%s}:%s }'
% (temp,writeFEvalsMaxPrec(bestalgdata[-1], 2)))
curlineHtml.append('<td><i>%s</i>:%s</td>\n'
% (temp, writeFEvalsMaxPrec(bestalgdata[-1], 2)))
else:
# write #fevals of the reference alg
for i in bestalgdata[:-1]:
curline.append(r'\multicolumn{2}{@{}c@{}}{%s \quad}'
% writeFEvalsMaxPrec(i, 2))
curlineHtml.append('<td>%s</td>\n' % writeFEvalsMaxPrec(i, 2))
curline.append(r'\multicolumn{2}{@{}c@{}|}{%s}'
% writeFEvalsMaxPrec(bestalgdata[-1], 2))
curlineHtml.append('<td>%s</td>\n' % writeFEvalsMaxPrec(bestalgdata[-1], 2))
tmp = bestalgentry.detEvals([targetf])[0][0]
tmp2 = numpy.sum(numpy.isnan(tmp) == False)
curline.append('%d' % (tmp2))
if tmp2 > 0:
curline.append('/%d' % len(tmp))
curlineHtml.append('<td>%d/%d</td>\n' % (tmp2, len(tmp)))
else:
curlineHtml.append('<td>%d</td>\n' % (tmp2))
table.append(curline[:])
tableHtml.extend(curlineHtml[:])
tableHtml.append('</tr>\n')
extraeol.append('')
rankdata0 = [] # never used
# generate all data from ranksum test
entries = []
ertdata = {}
for nb, dsList in enumerate((dictFunc0, dictFunc1)):
try:
entry = dsList[f][0] # take the first DataSet, there should be only one?
except KeyError:
warnings.warn('data missing for data set ' + str(nb) + ' and function ' + str(f))
print('*** Warning: data missing for data set ' + str(nb) + ' and function ' + str(f) + '***')
continue # TODO: problem here!
ertdata[nb] = entry.detERT(targets)
entries.append(entry)
for _t in ertdata.values():
for _tt in _t:
if _tt is None:
raise ValueError
if len(entries) < 2: # funcion not available for *both* algorithms
continue # TODO: check which one is missing and make sure that what is there is displayed properly in the following
testres0vs1 = significancetest(entries[0], entries[1], targets)
testresbestvs1 = significancetest(bestalgentry, entries[1], targets)
testresbestvs0 = significancetest(bestalgentry, entries[0], targets)
for nb, entry in enumerate(entries):
tableHtml.append('<tr>\n')
if nb == 0:
curline = [r'1:\:\algorithmAshort\hspace*{\fill}']
curlineHtml = ['<th>1: %s</th>\n' % alg0]
else:
curline = [r'2:\:\algorithmBshort\hspace*{\fill}']
curlineHtml = ['<th>2: %s</th>\n' % alg1]
#data = entry.detERT(targetsOfInterest)
dispersion = []
data = []
evals = entry.detEvals(targets)
for i in evals:
succ = (numpy.isnan(i) == False)
tmp = i.copy()
tmp[succ==False] = entry.maxevals[numpy.isnan(i)]
#set_trace()
data.append(toolsstats.sp(tmp, issuccessful=succ)[0])
#if not any(succ):
#set_trace()
if any(succ):
tmp2 = toolsstats.drawSP(tmp[succ], tmp[succ==False],
(10, 50, 90), samplesize)[0]
dispersion.append((tmp2[-1]-tmp2[0])/2.)
else:
dispersion.append(None)
if nb == 0:
assert not isinstance(data, numpy.ndarray)
data0 = data[:] # TODO: check if it is not an array, it's never used anyway?
for i, dati in enumerate(data):
z, p = testres0vs1[i] # TODO: there is something with the sign that I don't get
# assign significance flag, which is the -log10(p)
significance0vs1 = 0
if nb != 0:
z = -z # the test is symmetric
if nbtests * p < 0.05 and z > 0:
significance0vs1 = -int(numpy.ceil(numpy.log10(min([1.0, nbtests * p])))) # this is the larger the more significant
isBold = significance0vs1 > 0
alignment = 'c'
if i == len(data) - 1: # last element
alignment = 'c|'
if numpy.isinf(bestalgdata[i]): # if the 2009 best did not solve the problem
tmp = writeFEvalsMaxPrec(float(dati), 2)
if not numpy.isinf(dati):
tmpHtml = '<i>%s</i>' % (tmp)
tmp = r'\textit{%s}' % (tmp)
if isBold:
tmp = r'\textbf{%s}' % tmp
tmpHtml = '<b>%s</b>' % tmpHtml
if dispersion[i] and numpy.isfinite(dispersion[i]):
tmp += r'${\scriptscriptstyle (%s)}$' % writeFEvalsMaxPrec(dispersion[i], 1)
tableentry = (r'\multicolumn{2}{@{}%s@{}}{%s}'
% (alignment, tmp))
tableentryHtml = (' (%s)' % tmp)
else:
# Formatting
tmp = float(dati)/bestalgdata[i]
assert not numpy.isnan(tmp)
isscientific = False
if tmp >= 1000:
isscientific = True
tableentry = writeFEvals2(tmp, 2, isscientific=isscientific)
tableentry = writeFEvalsMaxPrec(tmp, 2)
tableentryHtml = writeFEvalsMaxPrec(tmp, 2)
if numpy.isinf(tmp) and i == len(data)-1:
tableentry = (tableentry
+ r'\textit{%s}' % writeFEvals2(numpy.median(entry.maxevals), 2))
tableentryHtml = (tableentryHtml
+ ' <i>%s</i>' % writeFEvals2(numpy.median(entry.maxevals), 2))
if isBold:
tableentry = r'\textbf{%s}' % tableentry
tableentryHtml = '<b>%s</b>' % tableentryHtml
elif 11 < 3 and significance0vs1 < 0: # cave: negative significance has no meaning anymore
tableentry = r'\textit{%s}' % tableentry
tableentryHtml = '<i>%s</i>' % tableentryHtml
if dispersion[i] and numpy.isfinite(dispersion[i]/bestalgdata[i]):
tableentry += r'${\scriptscriptstyle (%s)}$' % writeFEvalsMaxPrec(dispersion[i]/bestalgdata[i], 1)
tableentryHtml += ' (%s)' % writeFEvalsMaxPrec(dispersion[i]/bestalgdata[i], 1)
tableentry = (r'\multicolumn{2}{@{}%s@{}}{%s}'
% (alignment, tableentry))
elif tableentry.find('e') > -1 or (numpy.isinf(tmp) and i != len(data) - 1):
if isBold:
tableentry = r'\textbf{%s}' % tableentry
tableentryHtml = '<b>%s</b>' % tableentryHtml
elif 11 < 3 and significance0vs1 < 0:
tableentry = r'\textit{%s}' % tableentry
tableentryHtml = '<i>%s</i>' % tableentryHtml
if dispersion[i] and numpy.isfinite(dispersion[i]/bestalgdata[i]):
tableentry += r'${\scriptscriptstyle (%s)}$' % writeFEvalsMaxPrec(dispersion[i]/bestalgdata[i], 1)
tableentryHtml += ' (%s)' % writeFEvalsMaxPrec(dispersion[i]/bestalgdata[i], 1)
tableentry = (r'\multicolumn{2}{@{}%s@{}}{%s}'
% (alignment, tableentry))
else:
tmp = tableentry.split('.', 1)
tmpHtml = tableentryHtml.split('.', 1)
if isBold:
tmp = list(r'\textbf{%s}' % i for i in tmp)
tmpHtml = list('<b>%s</b>' % i for i in tmpHtml)
elif 11 < 3 and significance0vs1 < 0:
tmp = list(r'\textit{%s}' % i for i in tmp)
tmpHtml = list('<i>%s</i>' % i for i in tmpHtml)
tableentry = ' & .'.join(tmp)
tableentryHtml = '.'.join(tmpHtml)
if len(tmp) == 1:
tableentry += '&'
if dispersion[i] and numpy.isfinite(dispersion[i]/bestalgdata[i]):
tableentry += r'${\scriptscriptstyle (%s)}$' % writeFEvalsMaxPrec(dispersion[i]/bestalgdata[i], 1)
tableentryHtml += ' (%s)' % writeFEvalsMaxPrec(dispersion[i]/bestalgdata[i], 1)
superscript = ''
superscriptHtml = ''
if nb == 0:
z, p = testresbestvs0[i]
else:
z, p = testresbestvs1[i]
#The conditions are now that ERT < ERT_best
if ((nbtests * p) < 0.05 and dati - bestalgdata[i] < 0.
and z < 0.):
nbstars = -numpy.ceil(numpy.log10(nbtests * p))
#tmp = '\hspace{-.5ex}'.join(nbstars * [r'\star'])
if z > 0:
superscript = r'\uparrow' #* nbstars
superscriptHtml = '↑'
else:
superscript = r'\downarrow' #* nbstars
superscriptHtml = '↓'
# print z, linebest[i], line1
if nbstars > 1:
superscript += str(int(nbstars))
superscriptHtml += str(int(nbstars))
if superscript or significance0vs1:
s = ''
shtml = ''
if significance0vs1 > 0:
s = '\star'
shtml = '★'
if significance0vs1 > 1:
s += str(significance0vs1)
shtml += str(significance0vs1)
s = r'$^{' + s + superscript + r'}$'
shtml = '<sup>' + shtml + superscriptHtml + '</sup>'
if tableentry.endswith('}'):
tableentry = tableentry[:-1] + s + r'}'
else:
tableentry += s
tableentryHtml += shtml
tableentryHtml = tableentryHtml.replace('$\infty$', '∞')
curlineHtml.append('<td>%s</td>\n' % tableentryHtml)
curline.append(tableentry)
#curline.append(tableentry)
#if dispersion[i] is None or numpy.isinf(bestalgdata[i]):
#curline.append('')
#else:
#tmp = writeFEvalsMaxPrec(dispersion[i]/bestalgdata[i], 2)
#curline.append('(%s)' % tmp)
tmp = entry.evals[entry.evals[:, 0] <= targetf, 1:]
try:
tmp = tmp[0]
curline.append('%d' % numpy.sum(numpy.isnan(tmp) == False))
curlineHtml.append('<td>%d' % numpy.sum(numpy.isnan(tmp) == False))
except IndexError:
curline.append('%d' % 0)
curlineHtml.append('<td>%d' % 0)
curline.append('/%d' % entry.nbRuns())
curlineHtml.append('/%d</td>\n' % entry.nbRuns())
table.append(curline[:])
tableHtml.extend(curlineHtml[:])
tableHtml.append('</tr>\n')
extraeol.append('')
extraeol[-1] = r'\hline'
extraeol[-1] = ''
outputfile = os.path.join(outputdir, 'pptable2_%02dD%s.tex' % (d, info))
spec = r'@{}c@{}|' + '*{%d}{@{}r@{}@{}l@{}}' % len(targetsOfInterest) + '|@{}r@{}@{}l@{}'
res = r'\providecommand{\algorithmAshort}{%s}' % writeLabels(alg0) + '\n'
res += r'\providecommand{\algorithmBshort}{%s}' % writeLabels(alg1) + '\n'
# open(os.path.join(outputdir, 'bbob_pproc_commands.tex'), 'a').write(res)
#res += tableLaTeXStar(table, width=r'0.45\textwidth', spec=spec,
#extraeol=extraeol)
res += tableLaTeX(table, spec=spec, extraeol=extraeol)
f = open(outputfile, 'w')
f.write(res)
f.close()
res = ("").join(str(item) for item in tableHtml)
res = '<p><b>%d-D</b></p>\n<table>\n%s</table>\n' % (d, res)
filename = os.path.join(outputdir, genericsettings.two_algorithm_file_name + '.html')
lines = []
with open(filename) as infile:
for line in infile:
if '<!--pptable2Html-->' in line:
lines.append(res)
lines.append(line)
with open(filename, 'w') as outfile:
for line in lines:
outfile.write(line)
if verbose:
print "Table written in %s" % outputfile