def plot_csp_patterns_(all_patterns, sensor_names,
original_class_names=('Hand (R)', 'Hand (L)', 'Rest', 'Feet')):
"""Expects filterband x classpair x sensor x 2 (pattern).
Expects classpairs in original order, i.e.
Hand(R)/Hand(L), Hand(R)/Rest, Hand(R)/Feet,
Hand(L)/Rest, Hand(L)/Feet, Feet/Rest"""
fig = plt.figure(figsize=(12,2))
for i_class_pair in range(6):
i_wanted_class_pair, wanted_class_pair, reverse_filters = map_i_class_pair(i_class_pair)
pair_patterns = all_patterns[i_wanted_class_pair]
if reverse_filters:
pair_patterns = pair_patterns[:,::-1]
for i_sub_pattern in range(2):
pattern = pair_patterns[:,i_sub_pattern]
ax = plt.subplot(2,6, i_class_pair+(i_sub_pattern * 6)+1)
if i_sub_pattern == 0 and i_class_pair == 0:
scalp_line_width = 1
#ax.set_ylabel(u"{:.0f}—{:.0f} Hz".format(*filterbands[i_fb]))
else:
scalp_line_width = 0
ax_scalp(pattern,sensor_names, colormap=cm.PRGn, ax=ax,
vmin=-np.max(np.abs(pattern)), vmax=np.max(np.abs(pattern)),
scalp_line_width=scalp_line_width)
if i_sub_pattern == 0:
# reversefilters is 0 if not to be reversed and 1 if to be revrsed
ax.set_title(original_class_names[wanted_class_pair[reverse_filters]])
else:
ax.set_xlabel(original_class_names[wanted_class_pair[1-reverse_filters]])
fig.subplots_adjust(wspace=-0.7,hspace=-0)
add_colorbar_to_scalp_grid(fig,np.array(fig.axes),'', shrink=1)
return fig
def plot_csp_patterns_(all_patterns, sensor_names,
original_class_names=('Hand (R)', 'Hand (L)', 'Rest', 'Feet')):
"""Expects filterband x classpair x sensor x 2 (pattern).
Expects classpairs in original order, i.e.
Hand(R)/Hand(L), Hand(R)/Rest, Hand(R)/Feet,
Hand(L)/Rest, Hand(L)/Feet, Feet/Rest"""
fig = plt.figure(figsize=(12, 2))
for i_class_pair in range(6):
i_wanted_class_pair, wanted_class_pair, reverse_filters = map_i_class_pair(i_class_pair)
pair_patterns = all_patterns[i_wanted_class_pair]
if reverse_filters:
pair_patterns = pair_patterns[:, ::-1]
for i_sub_pattern in range(2):
pattern = pair_patterns[:, i_sub_pattern]
ax = plt.subplot(2, 6, i_class_pair + (i_sub_pattern * 6) + 1)
if i_sub_pattern == 0 and i_class_pair == 0:
scalp_line_width = 1
# ax.set_ylabel(u"{:.0f}–{:.0f} Hz".format(*filterbands[i_fb]))
else:
scalp_line_width = 0
ax_scalp(pattern, sensor_names, colormap=cm.PRGn, ax=ax,
vmin=-np.max(np.abs(pattern)), vmax=np.max(np.abs(pattern)),
scalp_line_width=scalp_line_width)
if i_sub_pattern == 0:
# reversefilters is 0 if not to be reversed and 1 if to be revrsed
ax.set_title(original_class_names[wanted_class_pair[reverse_filters]])
else:
ax.set_xlabel(original_class_names[wanted_class_pair[1 - reverse_filters]])
fig.subplots_adjust(wspace=-0.7, hspace=-0)
add_colorbar_to_scalp_grid(fig, np.array(fig.axes), '', shrink=1)
return fig
def plot_scalp_grid(data, sensor_names, scale_per_row=False,
scale_per_column=False,
scale_individually=False, figsize=None,
row_names=None, col_names=None,
vmin=None, vmax=None,
chan_pos_list=CHANNEL_10_20_APPROX,
colormap=cm.coolwarm,
fontsize=16):
"""
data: 3darray
freqs x classes x sensors
"""
assert np.sum([scale_per_row, scale_per_column, scale_individually]) < 2, (
"Can have only one way of scaling...")
if vmin is None:
assert vmax is None
max_abs_val = np.max(np.abs(data))
vmin = -max_abs_val
vmax = max_abs_val
n_rows = data.shape[0]
n_cols = data.shape[1]
if figsize is None:
figsize = (n_rows * 3, n_cols * 2)
fig, axes = plt.subplots(nrows=n_rows, ncols=n_cols, figsize=figsize)
for i_row in range(n_rows):
if scale_per_row:
max_abs_val = np.max(np.abs(data[i_row]))
vmin = -max_abs_val
vmax = max_abs_val
for i_col in xrange(n_cols):
this_data = data[i_row, i_col]
if scale_per_column:
max_abs_val = np.max(np.abs(data[:, i_col]))
vmin = -max_abs_val
vmax = max_abs_val
if scale_individually:
max_abs_val = np.max(np.abs(this_data))
vmin = -max_abs_val
vmax = max_abs_val
scalp_line_style = 'solid'
if i_row == 0 and i_col == 0:
scalp_line_width = 1
else:
scalp_line_width = 0
if n_rows > 1:
ax = axes[i_row][i_col]
else:
ax = axes[i_col]
ax_scalp(this_data, sensor_names, colormap=colormap, ax=ax,
scalp_line_width=scalp_line_width, scalp_line_style=scalp_line_style,
vmin=vmin, vmax=vmax, chan_pos_list=chan_pos_list)
if col_names is not None and i_row == 0:
ax.set_title(col_names[i_col], fontsize=fontsize)
if row_names is not None and i_col == 0:
ax.set_ylabel(row_names[i_row], fontsize=fontsize)
fig.subplots_adjust(hspace=-0.3, wspace=0.)
return fig, axes
def plot_scalp_grid(data, sensor_names, scale_per_row=False,
scale_per_column=False,
scale_individually=False, figsize=None,
row_names=None, col_names=None,
vmin=None, vmax=None,
colormap=cm.coolwarm,
fontsize=16):
"""
data: 3darray
freqs x classes x sensors
"""
assert np.sum([scale_per_row, scale_per_column, scale_individually]) < 2, (
"Can have only one way of scaling...")
if vmin is None:
assert vmax is None
max_abs_val = np.max(np.abs(data))
vmin = -max_abs_val
vmax = max_abs_val
n_rows = data.shape[0]
n_cols = data.shape[1]
if figsize is None:
figsize = (n_rows*3, n_cols*2)
fig, axes = plt.subplots(nrows=n_rows, ncols=n_cols,figsize=figsize)
for i_row in xrange(n_rows):
if scale_per_row:
max_abs_val = np.max(np.abs(data[i_row]))
vmin = -max_abs_val
vmax = max_abs_val
for i_col in xrange(n_cols):
this_data = data[i_row,i_col]
if scale_per_column:
max_abs_val = np.max(np.abs(data[:,i_col]))
vmin = -max_abs_val
vmax = max_abs_val
if scale_individually:
max_abs_val = np.max(np.abs(this_data))
vmin = -max_abs_val
vmax = max_abs_val
scalp_line_style = 'solid'
if i_row == 0 and i_col == 0:
scalp_line_width = 1
else:
scalp_line_width=0
if n_rows > 1:
ax = axes[i_row][i_col]
else:
ax = axes[i_col]
ax_scalp(this_data,sensor_names, colormap=colormap,ax=ax,
scalp_line_width=scalp_line_width, scalp_line_style=scalp_line_style,
vmin=vmin, vmax=vmax)
if col_names is not None and i_row == 0:
ax.set_title(col_names[i_col], fontsize=fontsize)
if row_names is not None and i_col == 0:
ax.set_ylabel(row_names[i_row], fontsize=fontsize)
fig.subplots_adjust(hspace=-0.3,wspace=0.)
return fig, axes
