def bxes(settings, anames, key, relative=False, base=None, **kwargs):
slist, plist, values = sts.settings_stats(settings, key, relative, base, anames)
vs = [y['values'] for y in values];
print vs
fig = plt.figure()
plt.boxplot(vs, sym='', whis=1)
locs, _ = plt.xticks()
plt.xticks(locs, [anames(s) for s in slist], rotation=0)
plt.grid(axis='y')
if 'title' in kwargs:
pylab.title(kwargs['title'])
if 'ylabel' in kwargs:
ylabel = kwargs['ylabel']
if base:
ylabel += " (\% over " + base + ")"
elif relative:
ylabel += " (\% over best per isntance)"
pylab.ylabel(ylabel)
if 'ylim' in kwargs:
pylab.ylim(kwargs['ylim'])
# Size adjustment
DefaultSize = fig.get_size_inches()
fig.subplots_adjust(top=0.9,bottom=0.23, right=0.95, left=0.12)
fig.set_size_inches(DefaultSize[0]*2./3, DefaultSize[1]*2./3)
def dataset_legend(settings, dataset_key, anames, key, relative=False, base=None, **kwargs):
fig = pylab.figure()
xs = sorted(set([s[dataset_key] for s in settings]))
x_settings = dict([(s,settings[s]) for s in settings if s[dataset_key] == xs[0]])
slist, plist, values = sts.settings_stats(x_settings, key, relative, base, anames)
pylab.title("legend " + dataset_key)
ax = fig.add_subplot(111)
ax.set_axis_off()
lsi = line_styles()
for s in slist:
ax.plot([0],[0], label=anames(s), linewidth=3, ls=lsi.next())
pylab.legend(ncol=len(slist), loc=10)
def dataset_lines(settings, dataset_key, anames, key, relative=False, base=None, **kwargs):
xs = sorted(set([s[dataset_key] for s in settings]))
ys = [[] for x in xs]
stat = 'mean'
for i, x in enumerate(xs):
x_settings = dict([(s,settings[s]) for s in settings if s[dataset_key] == x])
slist, plist, values = sts.settings_stats(x_settings, key, relative, base, anames)
stat_values = [v[stat] for v in values]
best = np.argmin(stat_values)
print "Best %s [%s]: %s (%f)" % (stat, x, anames(slist[best]), stat_values[best])
ys[i] = stat_values
fig = plt.figure()
ax = fig.add_subplot(111)
ys = np.array(ys)
lsi = line_styles()
ticklabels = xs
xs = range(len(xs))
for i in xrange(ys.shape[1]):
ax = plt.plot(xs, ys[:,i], label=anames(slist[i]), linewidth=3, \
ls=lsi.next(), marker='o', markersize=10)
plt.xticks(xs, ticklabels)
plt.grid(axis='y')
#plt.legend(loc='upper center', ncol=len(slist), bbox_to_anchor=(0, .1,1,1))
if 'title' in kwargs:
pylab.title(kwargs['title'])
if 'ylabel' in kwargs:
ylabel = stat + " " + kwargs['ylabel']
if base:
ylabel += " (\% over " + base + ")"
elif relative:
ylabel += " (\% over best per instance)"
pylab.ylabel(ylabel)
if 'xlabel' in kwargs:
pylab.xlabel(kwargs['xlabel'])
if 'ylim' in kwargs:
pylab.ylim(kwargs['ylim'])
def lines(settings, anames, key, xkey, relative=False, base=None, **kwargs):
slist, plist, values = sts.settings_stats(settings, key, relative, base, anames)
xs = [s[xkey] for s in slist]
fig = plt.figure()
plot_stats = ['mean']
for si, stat in enumerate(plot_stats):
stat_values = [v[stat] for v in values]
for vi, _ in enumerate(values):
print anames(slist[vi]), stat_values[vi]
best = np.argmin(stat_values)
print "Best %s: %s (%f)" % (stat, anames(slist[best]), stat_values[best])
ax = plt.subplot(len(plot_stats), 1, si+1)
ticklabels = xs
xs = range(len(xs))
plt.plot(xs, stat_values)
plt.xticks(xs, ticklabels)
plt.grid(axis='y')
if 'title' in kwargs:
pylab.title(kwargs['title'])
if 'ylabel' in kwargs:
ylabel = stat + " " + kwargs['ylabel']
if base:
ylabel += " (\% over " + base + ")"
elif relative:
ylabel += " (\% over best per instance)"
pylab.ylabel(ylabel)
if 'xlabel' in kwargs:
pylab.xlabel(kwargs['xlabel'])
if 'ylim' in kwargs:
pylab.ylim(kwargs['ylim'])
# Size adjustment
DefaultSize = fig.get_size_inches()
fig.subplots_adjust(top=0.9,bottom=0.23, right=0.95, left=0.12)
fig.set_size_inches(DefaultSize[0]*2./3, DefaultSize[1]*2./3)