def _plot(self):
col_dict = get_column_dict(self.collection, 'i', *self.metric_selector)
x = col_dict.pop('i')
self.figure.clf()
axes = self.figure.add_subplot(111)
if self.smooth > 0: x, = gaussConv(self.smooth, x )
# print x
for label, y in col_dict.items():
if self.smooth > 0: y, = gaussConv(self.smooth, y )
# print y
axes.plot(x,y, label=label)
axes.legend(loc='best')
axes.set_xlabel('iteration')
self.figure.canvas.draw()
def _plot(self):
col_dict = get_column_dict(self.collection, "i", *self.metric_selector)
x = col_dict.pop("i")
self.figure.clf()
axes = self.figure.add_subplot(111)
if self.smooth > 0:
x, = gaussConv(self.smooth, x)
# print x
for label, y in col_dict.items():
if self.smooth > 0:
y, = gaussConv(self.smooth, y)
# print y
axes.plot(x, y, label=label)
axes.legend(loc="best")
axes.set_xlabel("iteration")
self.figure.canvas.draw()
def plot_curve(collection, x_key, *y_key_list):
col_dict = get_column_dict(collection, *((x_key, ) + y_key_list))
x = col_dict.get(x_key)
x_, = gaussConv(1, x)
color_cycle = pp.gca()._get_lines.color_cycle
for y_key in y_key_list:
color = color_cycle.next()
y = col_dict.get(y_key)
y_, = gaussConv(1, y)
pp.plot(x, y, '.', color=color, markersize=2)
pp.plot(x_, y_, '-', label=y_key, color=color)
pp.xlabel(x_key)
pp.legend(loc='best')
def plot_curve(collection, x_key, *y_key_list):
col_dict = get_column_dict( collection, * ((x_key, ) + y_key_list) )
x = col_dict.get(x_key)
x_, = gaussConv(1,x )
color_cycle = pp.gca()._get_lines.color_cycle
for y_key in y_key_list:
color= color_cycle.next()
y = col_dict.get(y_key)
y_, = gaussConv(1,y )
pp.plot(x,y,'.', color=color,markersize=2)
pp.plot(x_,y_,'-', label=y_key, color=color)
pp.xlabel(x_key)
pp.legend(loc='best')
def plot_eval_info(plot, hp_info, y_keys, perm=None):
y_dict = get_column_dict(hp_info.trace.db.eval_info, 'hp_id', *y_keys)
idx = hp_info.map_hp_id_list(y_dict.pop('hp_id'))
# add the agnostic bayes distribution the the list of traces
idx_list, distr_list = get_column_list(hp_info.trace.db.predict, 'i',
'prob')
distr = distr_list[np.argmax(
idx_list)] # extract the last computed distribution
y_dict['AB probability'] = unpack_prob(distr, hp_info, len(idx))
if len(idx) == 0:
print 'no results yet'
return
gp = MyGP(mcmc_iters=0, noiseless=False)
gp.set_hypers(hp_info.chooser_state)
for key in y_keys:
y_dict[key] = np.array(y_dict[key])
# print '%s.shape:'%y_key, y.shape
X = hp_info.unit_grid[idx, :]
hp_keys = hp_info.hp_keys
print hp_keys
if perm is not None:
X = X[:, perm]
hp_keys = [hp_keys[i] for i in perm]
hp_keys = [clean_hp_name(hp_key) for hp_key in hp_keys]
print hp_keys
plot.set_info(X, y_dict, 'val.risk', hp_keys, hp_info.hp_space.var_list,
gp)
def plot_eval_info( plot, hp_info, y_keys, perm = None ):
y_dict = get_column_dict( hp_info.trace.db.eval_info, 'hp_id', *y_keys )
idx = hp_info.map_hp_id_list(y_dict.pop('hp_id'))
# add the agnostic bayes distribution the the list of traces
idx_list, distr_list = get_column_list( hp_info.trace.db.predict, 'i', 'prob' )
distr = distr_list[ np.argmax(idx_list) ] # extract the last computed distribution
y_dict['AB probability'] = unpack_prob( distr, hp_info, len(idx))
if len(idx) == 0:
print 'no results yet'
return
gp = MyGP(mcmc_iters=0, noiseless=False)
gp.set_hypers(hp_info.chooser_state)
for key in y_keys:
y_dict[key] = np.array(y_dict[key])
# print '%s.shape:'%y_key, y.shape
X = hp_info.unit_grid[idx,:]
hp_keys = hp_info.hp_keys
print hp_keys
if perm is not None:
X = X[:,perm]
hp_keys = [hp_keys[i] for i in perm ]
hp_keys = [ clean_hp_name(hp_key) for hp_key in hp_keys ]
print hp_keys
plot.set_info(X, y_dict, 'val.risk',hp_keys, hp_info.hp_space.var_list, gp)
