spin on a/o, colorbar on keypress
This commit is contained in:
parent
63be80aa92
commit
2cfa3ca415
@ -15,94 +15,169 @@ from fluents import plots
|
||||
from scipy import dot,sum,diag,arange,log,mean,newaxis,sqrt
|
||||
from matplotlib import cm
|
||||
import pylab as PB
|
||||
import gtk
|
||||
|
||||
class PcaScorePlot(plots.ScatterPlot):
|
||||
"""PCA Score plot"""
|
||||
def __init__(self, model, absi=0, ordi=1):
|
||||
self._T = model.model['T']
|
||||
dataset_1 = model.as_dataset('T')
|
||||
dataset_2 = dataset_1
|
||||
|
||||
class BlmScatterPlot(plots.ScatterPlot):
|
||||
"""Scatter plot used for scores and loadings in bilinear models."""
|
||||
|
||||
def __init__(self, title, model, absi=0, ordi=1, part_name='T', color_by=None):
|
||||
if model.model.has_key(part_name)!=True:
|
||||
raise ValueError("Model part: %s not found in model" %mod_param)
|
||||
self._T = model.model[part_name]
|
||||
if self._T.shape[1]==1:
|
||||
logger.log('notice', 'Scores have only one component')
|
||||
absi= ordi = 0
|
||||
self._absi = absi
|
||||
self._ordi = ordi
|
||||
self._colorbar = None
|
||||
dataset_1 = model.as_dataset(part_name)
|
||||
id_dim = dataset_1.get_dim_name(0)
|
||||
sel_dim = dataset_1.get_dim_name(1)
|
||||
id_1, = dataset_1.get_identifiers(sel_dim, [absi])
|
||||
id_2, = dataset_1.get_identifiers(sel_dim, [ordi])
|
||||
plots.ScatterPlot.__init__(self, dataset_1, dataset_2, id_dim, sel_dim, id_1, id_2 ,c='b' ,s=40 , name='pca-scores')
|
||||
|
||||
def set_absicca(self, n):
|
||||
self.xaxis_data = self._T[:,n]
|
||||
col = 'b'
|
||||
if model.model.has_key(color_by):
|
||||
col = model.model[color_by].ravel()
|
||||
|
||||
plots.ScatterPlot.__init__(self, dataset_1, dataset_1, id_dim, sel_dim, id_1, id_2 ,c=col ,s=40 , name=title)
|
||||
|
||||
def set_ordinate(self, n):
|
||||
self.yaxis_data = self._T[:,n]
|
||||
self.add_pc_spin_buttons(self._T.shape[1], absi, ordi)
|
||||
self._key_press = self.canvas.mpl_connect(
|
||||
'key_press_event', self._on_key_press)
|
||||
|
||||
class PcaLoadingPlot(plots.ScatterPlot):
|
||||
"""PCA Loading plot"""
|
||||
def __init__(self, model, absi=0, ordi=1):
|
||||
self._P = model.model['P']
|
||||
dataset_1 = model.as_dataset('P')
|
||||
dataset_2 = dataset_1
|
||||
id_dim = dataset_1.get_dim_name(0)
|
||||
sel_dim = dataset_1.get_dim_name(1)
|
||||
id_1, = dataset_1.get_identifiers(sel_dim, [absi])
|
||||
id_2, = dataset_1.get_identifiers(sel_dim, [ordi])
|
||||
if model.model.has_key('p_tsq'):
|
||||
col = model.model['p_tsq'].ravel()
|
||||
col = normalise(col)
|
||||
else:
|
||||
col = 'g'
|
||||
plots.ScatterPlot.__init__(self, dataset_1, dataset_2, id_dim, sel_dim, id_1, id_2,c=col,s=20, name='pls-loadings')
|
||||
def _on_key_press(self, event):
|
||||
if event.key=='c':
|
||||
self.toggle_colorbar()
|
||||
|
||||
def set_absicca(self, n):
|
||||
self.xaxis_data = self._P[:,n]
|
||||
def set_facecolor(self, colors):
|
||||
"""Set patch facecolors.
|
||||
"""
|
||||
pass
|
||||
|
||||
def set_ordinate(self, n):
|
||||
self.yaxis_data = self._P[:,n]
|
||||
|
||||
class PlsScorePlot(plots.ScatterPlot):
|
||||
"""PLS Score plot"""
|
||||
def __init__(self, model, absi=0, ordi=1):
|
||||
self._T = model.model['T']
|
||||
dataset_1 = model.as_dataset('T')
|
||||
dataset_2 = dataset_1
|
||||
id_dim = dataset_1.get_dim_name(0)
|
||||
sel_dim = dataset_1.get_dim_name(1)
|
||||
id_1, = dataset_1.get_identifiers(sel_dim, [absi])
|
||||
id_2, = dataset_1.get_identifiers(sel_dim, [ordi])
|
||||
|
||||
plots.ScatterPlot.__init__(self, dataset_1, dataset_2,
|
||||
id_dim, sel_dim, id_1, id_2 ,
|
||||
c='b' ,s=40 , name='pls-scores')
|
||||
|
||||
def set_absicca(self, n):
|
||||
self.xaxis_data = self._T[:,n]
|
||||
def set_alphas(self, alphas):
|
||||
"""Set alpha channel for all patches."""
|
||||
pass
|
||||
|
||||
def set_ordinate(self, n):
|
||||
self.yaxis_data = self._T[:,n]
|
||||
def set_sizes(self, sizes):
|
||||
"""Set patch sizes."""
|
||||
pass
|
||||
|
||||
|
||||
class PlsLoadingPlot(plots.ScatterPlot):
|
||||
"""PLS Loading plot"""
|
||||
def __init__(self, model, absi=0, ordi=1):
|
||||
self._P = model.model['P']
|
||||
dataset_1 = model.as_dataset('P')
|
||||
dataset_2 = dataset_1
|
||||
id_dim = dataset_1.get_dim_name(0)
|
||||
sel_dim = dataset_1.get_dim_name(1)
|
||||
id_1, = dataset_1.get_identifiers(sel_dim, [absi])
|
||||
id_2, = dataset_1.get_identifiers(sel_dim, [ordi])
|
||||
if model.model.has_key('w_tsq'):
|
||||
col = model.model['w_tsq'].ravel()
|
||||
col = normalise(col)
|
||||
def toggle_colorbar(self):
|
||||
if self._colorbar==None:
|
||||
if self.sc._A!=None: # we need colormapping
|
||||
# get axes original position
|
||||
self._ax_last_pos = self.ax.get_position()
|
||||
self._colorbar = self.fig.colorbar(self.sc)
|
||||
self._colorbar.draw_all()
|
||||
self.canvas.draw()
|
||||
else:
|
||||
col = 'g'
|
||||
plots.ScatterPlot.__init__(self, dataset_1, dataset_2,
|
||||
id_dim, sel_dim, id_1, id_2,
|
||||
c=col, s=20, name='loadings')
|
||||
# remove colorbar
|
||||
# remove, axes, observers, colorbar instance, and restore viewlims
|
||||
cb, ax = self.sc.colorbar
|
||||
self.fig.delaxes(ax)
|
||||
self.sc.observers = [obs for obs in self.sc.observers if obs !=self._colorbar]
|
||||
self._colorbar = None
|
||||
self.sc.colorbar = None
|
||||
self.ax.set_position(self._ax_last_pos)
|
||||
self.canvas.draw()
|
||||
|
||||
def add_pc_spin_buttons(self, amax, absi, ordi):
|
||||
sb_a = gtk.SpinButton(climb_rate=1)
|
||||
sb_a.set_range(1, amax)
|
||||
sb_a.set_value(absi)
|
||||
sb_a.set_increments(1, 5)
|
||||
sb_a.connect('value_changed', self.set_absicca)
|
||||
sb_o = gtk.SpinButton(climb_rate=1)
|
||||
sb_o.set_range(1, amax)
|
||||
sb_o.set_value(ordi)
|
||||
sb_o.set_increments(1, 5)
|
||||
sb_o.connect('value_changed', self.set_ordinate)
|
||||
hbox = gtk.HBox()
|
||||
gtk_label_a = gtk.Label("A:")
|
||||
gtk_label_o = gtk.Label(" O:")
|
||||
toolitem = gtk.ToolItem()
|
||||
toolitem.set_expand(False)
|
||||
toolitem.set_border_width(2)
|
||||
toolitem.add(hbox)
|
||||
hbox.pack_start(gtk_label_a)
|
||||
hbox.pack_start(sb_a)
|
||||
hbox.pack_start(gtk_label_o)
|
||||
hbox.pack_start(sb_o)
|
||||
self._toolbar.insert(toolitem, -1)
|
||||
toolitem.set_tooltip(self._toolbar.tooltips, "Set Principal component")
|
||||
self._toolbar.show_all() #do i need this?
|
||||
|
||||
def set_absicca(self, sb):
|
||||
self._absi = sb.get_value_as_int() - 1
|
||||
xy = self._T[:,[self._absi, self._ordi]]
|
||||
self.xaxis_data = xy[:,0]
|
||||
self.yaxis_data = xy[:,1]
|
||||
self.sc._offsets = xy
|
||||
if self.use_blit==True:
|
||||
self.canvas.restore_region(self._clean_bck)
|
||||
self.ax.draw_artist(self.sc)
|
||||
self.canvas.blit()
|
||||
else:
|
||||
self.canvas.draw_idle()
|
||||
|
||||
def set_absicca(self, n):
|
||||
self.xaxis_data = self._P[:,n]
|
||||
def set_ordinate(self, sb):
|
||||
self._ordi = sb.get_value_as_int() - 1
|
||||
xy = self._T[:,[self._absi, self._ordi]]
|
||||
self.xaxis_data = xy[:,0]
|
||||
self.yaxis_data = xy[:,1]
|
||||
self.sc._offsets = xy
|
||||
if self.use_blit==True:
|
||||
self.canvas.restore_region(self._clean_bck)
|
||||
self.ax.draw_artist(self.sc)
|
||||
self.canvas.blit()
|
||||
else:
|
||||
self.canvas.draw_idle()
|
||||
|
||||
def show_labels(self, index=None):
|
||||
if self._text_labels == None:
|
||||
x = self.xaxis_data
|
||||
y = self.yaxis_data
|
||||
self._text_labels = {}
|
||||
for name, n in self.dataset_1[self.current_dim].items():
|
||||
txt = self.ax.text(x[n],y[n], name)
|
||||
txt.set_visible(False)
|
||||
self._text_labels[n] = txt
|
||||
if index!=None:
|
||||
self.hide_labels()
|
||||
for indx,txt in self._text_labels.items():
|
||||
if indx in index:
|
||||
txt.set_visible(True)
|
||||
self.canvas.draw()
|
||||
|
||||
def hide_labels(self):
|
||||
for txt in self._text_labels.values():
|
||||
txt.set_visible(False)
|
||||
self.canvas.draw()
|
||||
|
||||
class PcaScorePlot(BlmScatterPlot):
|
||||
def __init__(self, model, absi=0, ordi=1):
|
||||
title = "Pca scores (%s)" %model._dataset['X'].get_name()
|
||||
BlmScatterPlot.__init__(self, title, model, absi, ordi, 'T')
|
||||
|
||||
|
||||
class PcaLoadingPlot(BlmScatterPlot):
|
||||
def __init__(self, model, absi=0, ordi=1):
|
||||
title = "Pca loadings (%s)" %model._dataset['X'].get_name()
|
||||
BlmScatterPlot.__init__(self, title, model, absi, ordi, part_name='P', color_by='p_tsq')
|
||||
|
||||
def set_ordinate(self, n):
|
||||
self.yaxis_data = self._T[:,n]
|
||||
|
||||
class PlsScorePlot(BlmScatterPlot):
|
||||
def __init__(self, model, absi=0, ordi=1):
|
||||
title = "Pls scores (%s)" %model._dataset['X'].get_name()
|
||||
BlmScatterPlot.__init__(self, title, model, absi, ordi, 'T')
|
||||
|
||||
|
||||
class PlsLoadingPlot(BlmScatterPlot):
|
||||
def __init__(self, model, absi=0, ordi=1):
|
||||
title = "Pca loadings (%s)" %model._dataset['X'].get_name()
|
||||
BlmScatterPlot.__init__(self, title, model, absi, ordi, part_name='P', color_by='w_tsq')
|
||||
|
||||
|
||||
class LineViewXc(plots.LineViewPlot):
|
||||
@ -127,7 +202,7 @@ class PlsQvalScatter(plots.ScatterPlot):
|
||||
"""
|
||||
def __init__(self, model, pc=0):
|
||||
if not model.model.has_key('w_tsq'):
|
||||
return
|
||||
return None
|
||||
self._W = model.model['P']
|
||||
dataset_1 = model.as_dataset('P')
|
||||
dataset_2 = model.as_dataset('w_tsq')
|
||||
@ -146,13 +221,14 @@ class PlsQvalScatter(plots.ScatterPlot):
|
||||
c=col, s=20, sel_dim_2=sel_dim_2,
|
||||
name='Load Volcano')
|
||||
|
||||
|
||||
class PredictionErrorPlot(plots.Plot):
|
||||
"""A boxplot of prediction error vs. comp. number.
|
||||
"""
|
||||
def __init__(self, model, name="Pred. Err."):
|
||||
def __init__(self, model, name="Prediction Error"):
|
||||
if not model.model.has_key('sep'):
|
||||
logger.log('notice', 'Model has no calculations of sep')
|
||||
return
|
||||
return None
|
||||
plots.Plot.__init__(self, name)
|
||||
self._frozen = True
|
||||
self.current_dim = 'johndoe'
|
||||
@ -180,6 +256,15 @@ class InfluencePlot(plots.ScatterPlot):
|
||||
pass
|
||||
|
||||
|
||||
class RMSEPPlot(plots.BarPlot):
|
||||
def __init__(self, model, name="RMSEP"):
|
||||
if not model.model.has_key('rmsep'):
|
||||
logger.log('notice', 'Model has no calculations of sep')
|
||||
return
|
||||
dataset = model.as_dataset('rmsep')
|
||||
plots.BarPlot.__init__(self, dataset, name=name)
|
||||
|
||||
|
||||
def normalise(x):
|
||||
"""Scale vector x to [0,1]
|
||||
"""
|
||||
|
Reference in New Issue
Block a user