braindecode 0.1.5.4__tar.gz → 0.1.6__tar.gz
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- {Braindecode-0.1.5.4 → Braindecode-0.1.6}/Braindecode.egg-info/PKG-INFO +1 -1
- {Braindecode-0.1.5.4 → Braindecode-0.1.6}/Braindecode.egg-info/SOURCES.txt +2 -0
- {Braindecode-0.1.5.4 → Braindecode-0.1.6}/PKG-INFO +1 -1
- Braindecode-0.1.6/braindecode/datasets/sensor_positions.py +217 -0
- {Braindecode-0.1.5.4 → Braindecode-0.1.6}/braindecode/util.py +95 -0
- {Braindecode-0.1.5.4 → Braindecode-0.1.6}/docs/conf.py +2 -2
- {Braindecode-0.1.5.4 → Braindecode-0.1.6}/docs/index.rst +1 -0
- {Braindecode-0.1.5.4 → Braindecode-0.1.6}/docs/notebooks/Cropped_Decoding.ipynb +4 -4
- {Braindecode-0.1.5.4 → Braindecode-0.1.6}/docs/notebooks/TrialWise_Decoding.ipynb +9 -4
- Braindecode-0.1.6/docs/notebooks/visualization/Perturbation.ipynb +1840 -0
- {Braindecode-0.1.5.4 → Braindecode-0.1.6}/docs/source/braindecode.datasets.rst +8 -0
- {Braindecode-0.1.5.4 → Braindecode-0.1.6}/setup.py +1 -1
- {Braindecode-0.1.5.4 → Braindecode-0.1.6}/Braindecode.egg-info/dependency_links.txt +0 -0
- {Braindecode-0.1.5.4 → Braindecode-0.1.6}/Braindecode.egg-info/not-zip-safe +0 -0
- {Braindecode-0.1.5.4 → Braindecode-0.1.6}/Braindecode.egg-info/requires.txt +0 -0
- {Braindecode-0.1.5.4 → Braindecode-0.1.6}/Braindecode.egg-info/top_level.txt +0 -0
- {Braindecode-0.1.5.4 → Braindecode-0.1.6}/LICENSE.txt +0 -0
- {Braindecode-0.1.5.4 → Braindecode-0.1.6}/MANIFEST.in +0 -0
- {Braindecode-0.1.5.4 → Braindecode-0.1.6}/README.rst +0 -0
- {Braindecode-0.1.5.4 → Braindecode-0.1.6}/braindecode/__init__.py +0 -0
- {Braindecode-0.1.5.4 → Braindecode-0.1.6}/braindecode/datasets/__init__.py +0 -0
- {Braindecode-0.1.5.4 → Braindecode-0.1.6}/braindecode/datasets/bbci.py +0 -0
- {Braindecode-0.1.5.4 → Braindecode-0.1.6}/braindecode/datasets/bcic_iv_2a.py +0 -0
- {Braindecode-0.1.5.4 → Braindecode-0.1.6}/braindecode/datasets/multiple.py +0 -0
- {Braindecode-0.1.5.4 → Braindecode-0.1.6}/braindecode/datautil/__init__.py +0 -0
- {Braindecode-0.1.5.4 → Braindecode-0.1.6}/braindecode/datautil/iterators.py +0 -0
- {Braindecode-0.1.5.4 → Braindecode-0.1.6}/braindecode/datautil/signal_target.py +0 -0
- {Braindecode-0.1.5.4 → Braindecode-0.1.6}/braindecode/datautil/signalproc.py +0 -0
- {Braindecode-0.1.5.4 → Braindecode-0.1.6}/braindecode/datautil/splitters.py +0 -0
- {Braindecode-0.1.5.4 → Braindecode-0.1.6}/braindecode/datautil/trial_segment.py +0 -0
- {Braindecode-0.1.5.4 → Braindecode-0.1.6}/braindecode/datautil/util.py +0 -0
- {Braindecode-0.1.5.4 → Braindecode-0.1.6}/braindecode/experiments/__init__.py +0 -0
- {Braindecode-0.1.5.4 → Braindecode-0.1.6}/braindecode/experiments/experiment.py +0 -0
- {Braindecode-0.1.5.4 → Braindecode-0.1.6}/braindecode/experiments/monitors.py +0 -0
- {Braindecode-0.1.5.4 → Braindecode-0.1.6}/braindecode/experiments/stopcriteria.py +0 -0
- {Braindecode-0.1.5.4 → Braindecode-0.1.6}/braindecode/mne_ext/__init__.py +0 -0
- {Braindecode-0.1.5.4 → Braindecode-0.1.6}/braindecode/mne_ext/signalproc.py +0 -0
- {Braindecode-0.1.5.4 → Braindecode-0.1.6}/braindecode/models/__init__.py +0 -0
- {Braindecode-0.1.5.4 → Braindecode-0.1.6}/braindecode/models/deep4.py +0 -0
- {Braindecode-0.1.5.4 → Braindecode-0.1.6}/braindecode/models/eegnet.py +0 -0
- {Braindecode-0.1.5.4 → Braindecode-0.1.6}/braindecode/models/shallow_fbcsp.py +0 -0
- {Braindecode-0.1.5.4 → Braindecode-0.1.6}/braindecode/models/util.py +0 -0
- {Braindecode-0.1.5.4 → Braindecode-0.1.6}/braindecode/torch_ext/__init__.py +0 -0
- {Braindecode-0.1.5.4 → Braindecode-0.1.6}/braindecode/torch_ext/constraints.py +0 -0
- {Braindecode-0.1.5.4 → Braindecode-0.1.6}/braindecode/torch_ext/functions.py +0 -0
- {Braindecode-0.1.5.4 → Braindecode-0.1.6}/braindecode/torch_ext/init.py +0 -0
- {Braindecode-0.1.5.4 → Braindecode-0.1.6}/braindecode/torch_ext/losses.py +0 -0
- {Braindecode-0.1.5.4 → Braindecode-0.1.6}/braindecode/torch_ext/modules.py +0 -0
- {Braindecode-0.1.5.4 → Braindecode-0.1.6}/braindecode/torch_ext/util.py +0 -0
- {Braindecode-0.1.5.4 → Braindecode-0.1.6}/docs/Makefile +0 -0
- {Braindecode-0.1.5.4 → Braindecode-0.1.6}/docs/notebooks/BBCI_Data.ipynb +0 -0
- {Braindecode-0.1.5.4 → Braindecode-0.1.6}/docs/notebooks/BBCI_Data_Start_Stop.ipynb +0 -0
- {Braindecode-0.1.5.4 → Braindecode-0.1.6}/docs/notebooks/BBCI_Data_Trialwise.ipynb +0 -0
- {Braindecode-0.1.5.4 → Braindecode-0.1.6}/docs/notebooks/Experiment_Class.ipynb +0 -0
- {Braindecode-0.1.5.4 → Braindecode-0.1.6}/docs/source/braindecode.datautil.rst +0 -0
- {Braindecode-0.1.5.4 → Braindecode-0.1.6}/docs/source/braindecode.experiments.rst +0 -0
- {Braindecode-0.1.5.4 → Braindecode-0.1.6}/docs/source/braindecode.mne_ext.rst +0 -0
- {Braindecode-0.1.5.4 → Braindecode-0.1.6}/docs/source/braindecode.models.rst +0 -0
- {Braindecode-0.1.5.4 → Braindecode-0.1.6}/docs/source/braindecode.rst +0 -0
- {Braindecode-0.1.5.4 → Braindecode-0.1.6}/docs/source/braindecode.torch_ext.rst +0 -0
- {Braindecode-0.1.5.4 → Braindecode-0.1.6}/docs/source/modules.rst +0 -0
- {Braindecode-0.1.5.4 → Braindecode-0.1.6}/setup.cfg +0 -0
|
@@ -14,6 +14,7 @@ braindecode/datasets/__init__.py
|
|
|
14
14
|
braindecode/datasets/bbci.py
|
|
15
15
|
braindecode/datasets/bcic_iv_2a.py
|
|
16
16
|
braindecode/datasets/multiple.py
|
|
17
|
+
braindecode/datasets/sensor_positions.py
|
|
17
18
|
braindecode/datautil/__init__.py
|
|
18
19
|
braindecode/datautil/iterators.py
|
|
19
20
|
braindecode/datautil/signal_target.py
|
|
@@ -48,6 +49,7 @@ docs/notebooks/BBCI_Data_Trialwise.ipynb
|
|
|
48
49
|
docs/notebooks/Cropped_Decoding.ipynb
|
|
49
50
|
docs/notebooks/Experiment_Class.ipynb
|
|
50
51
|
docs/notebooks/TrialWise_Decoding.ipynb
|
|
52
|
+
docs/notebooks/visualization/Perturbation.ipynb
|
|
51
53
|
docs/source/braindecode.datasets.rst
|
|
52
54
|
docs/source/braindecode.datautil.rst
|
|
53
55
|
docs/source/braindecode.experiments.rst
|
|
@@ -0,0 +1,217 @@
|
|
|
1
|
+
import numpy as np
|
|
2
|
+
import math
|
|
3
|
+
|
|
4
|
+
|
|
5
|
+
CHANNEL_10_20_APPROX = ('angle',
|
|
6
|
+
('Fpz',(0.000, 4.000)),
|
|
7
|
+
('Fp1',(-3.500, 3.500)),
|
|
8
|
+
('Fp2',(3.500, 3.500)),
|
|
9
|
+
('AFp3h',(-1.000, 3.500)),
|
|
10
|
+
('AFp4h',(1.000, 3.500)),
|
|
11
|
+
('AF7',(-4.000, 3.000)),
|
|
12
|
+
('AF3',(-2.000, 3.000)),
|
|
13
|
+
('AFz',(0.000, 3.000)),
|
|
14
|
+
('AF4',(2.000, 3.000)),
|
|
15
|
+
('AF8',(4.000, 3.000)),
|
|
16
|
+
('AFF5h',(-2.500, 2.500)),
|
|
17
|
+
('AFF1',(-0.500, 2.500)),
|
|
18
|
+
('AFF2',(0.500, 2.500)),
|
|
19
|
+
('AFF6h',(2.500, 2.500)),
|
|
20
|
+
('F7',(-4.000, 2.000)),
|
|
21
|
+
('F5',(-3.000, 2.000)),
|
|
22
|
+
('F3',(-2.000, 2.000)),
|
|
23
|
+
('F1',(-1.000, 2.000)),
|
|
24
|
+
('Fz',(0.000, 2.000)),
|
|
25
|
+
('F2',(1.000, 2.000)),
|
|
26
|
+
('F4',(2.000, 2.000)),
|
|
27
|
+
('F6',(3.000, 2.000)),
|
|
28
|
+
('F8',(4.000, 2.000)),
|
|
29
|
+
('FFT7h',(-3.500, 1.500)),
|
|
30
|
+
('FFC5h',(-2.500, 1.500)),
|
|
31
|
+
('FFC3h',(-1.500, 1.500)),
|
|
32
|
+
('FFC1h',(-0.500, 1.500)),
|
|
33
|
+
('FFC2h',(0.500, 1.500)),
|
|
34
|
+
('FFC4h',(1.500, 1.500)),
|
|
35
|
+
('FFC6h',(2.500, 1.500)),
|
|
36
|
+
('FFT8h',(3.500, 1.500)),
|
|
37
|
+
('FT9',(-5.000, 1.000)),
|
|
38
|
+
('FT7',(-4.000, 1.000)),
|
|
39
|
+
('FC5',(-3.000, 1.000)),
|
|
40
|
+
('FC3',(-2.000, 1.000)),
|
|
41
|
+
('FC1',(-1.000, 1.000)),
|
|
42
|
+
('FCz',(0.000, 1.000)),
|
|
43
|
+
('FC2',(1.000, 1.000)),
|
|
44
|
+
('FC4',(2.000, 1.000)),
|
|
45
|
+
('FC6',(3.000, 1.000)),
|
|
46
|
+
('FT8',(4.000, 1.000)),
|
|
47
|
+
('FT10',(5.000, 1.000)),
|
|
48
|
+
('FTT9h',(-4.500, 0.500)),
|
|
49
|
+
('FTT7h',(-3.500, 0.500)),
|
|
50
|
+
('FCC5h',(-2.500, 0.500)),
|
|
51
|
+
('FCC3h',(-1.500, 0.500)),
|
|
52
|
+
('FCC1h',(-0.500, 0.500)),
|
|
53
|
+
('FCC2h',(0.500, 0.500)),
|
|
54
|
+
('FCC4h',(1.500, 0.500)),
|
|
55
|
+
('FCC6h',(2.500, 0.500)),
|
|
56
|
+
('FTT8h',(3.500, 0.500)),
|
|
57
|
+
('FTT10h',(4.500, 0.500)),
|
|
58
|
+
('M1',(-5.000, 0.000)),
|
|
59
|
+
# notsure if correct:
|
|
60
|
+
('T9', (-4.500, 0.000)),
|
|
61
|
+
('T7',(-4.000, 0.000)),
|
|
62
|
+
('C5',(-3.000, 0.000)),
|
|
63
|
+
('C3',(-2.000, 0.000)),
|
|
64
|
+
('C1',(-1.000, 0.000)),
|
|
65
|
+
('Cz',(0.000, 0.000)),
|
|
66
|
+
('C2',(1.000, 0.000)),
|
|
67
|
+
('C4',(2.000, 0.000)),
|
|
68
|
+
('C6',(3.000, 0.000)),
|
|
69
|
+
('T8',(4.000, 0.000)),
|
|
70
|
+
('T10', (4.500, 0.000)),
|
|
71
|
+
('M2',(5.000, 0.000)),
|
|
72
|
+
('TTP7h',(-3.500, -0.500)),
|
|
73
|
+
('CCP5h',(-2.500, -0.500)),
|
|
74
|
+
('CCP3h',(-1.500, -0.500)),
|
|
75
|
+
('CCP1h',(-0.500, -0.500)),
|
|
76
|
+
('CCP2h',(0.500, -0.500)),
|
|
77
|
+
('CCP4h',(1.500, -0.500)),
|
|
78
|
+
('CCP6h',(2.500, -0.500)),
|
|
79
|
+
('TTP8h',(3.500, -0.500)),
|
|
80
|
+
('TP7',(-4.000, -1.000)),
|
|
81
|
+
('CP5',(-3.000, -1.000)),
|
|
82
|
+
('CP3',(-2.000, -1.000)),
|
|
83
|
+
('CP1',(-1.000, -1.000)),
|
|
84
|
+
('CPz',(0.000, -1.000)),
|
|
85
|
+
('CP2',(1.000, -1.000)),
|
|
86
|
+
('CP4',(2.000, -1.000)),
|
|
87
|
+
('CP6',(3.000, -1.000)),
|
|
88
|
+
('TP8',(4.000, -1.000)),
|
|
89
|
+
('TPP9h',(-4.500, -1.500)),
|
|
90
|
+
('TPP7h',(-3.500, -1.500)),
|
|
91
|
+
('CPP5h',(-2.500, -1.500)),
|
|
92
|
+
('CPP3h',(-1.500, -1.500)),
|
|
93
|
+
('CPP1h',(-0.500, -1.500)),
|
|
94
|
+
('CPP2h',(0.500, -1.500)),
|
|
95
|
+
('CPP4h',(1.500, -1.500)),
|
|
96
|
+
('CPP6h',(2.500, -1.500)),
|
|
97
|
+
('TPP8h',(3.500, -1.500)),
|
|
98
|
+
('TPP10h',(4.500, -1.500)),
|
|
99
|
+
('P9',(-5.000, -2.000)),
|
|
100
|
+
('P7',(-4.000, -2.000)),
|
|
101
|
+
('P5',(-3.000, -2.000)),
|
|
102
|
+
('P3',(-2.000, -2.000)),
|
|
103
|
+
('P1',(-1.000, -2.000)),
|
|
104
|
+
('Pz',(0.000, -2.000)),
|
|
105
|
+
('P2',(1.000, -2.000)),
|
|
106
|
+
('P4',(2.000, -2.000)),
|
|
107
|
+
('P6',(3.000, -2.000)),
|
|
108
|
+
('P8',(4.000, -2.000)),
|
|
109
|
+
('P10',(5.000, -2.000)),
|
|
110
|
+
('PPO9h',(-4.500, -2.500)),
|
|
111
|
+
('PPO5h',(-3.000, -2.500)),
|
|
112
|
+
('PPO1',(-0.650, -2.500)),
|
|
113
|
+
('PPO2',(0.650, -2.500)),
|
|
114
|
+
('PPO6h',(3.000, -2.500)),
|
|
115
|
+
('PPO10h',(4.500, -2.500)),
|
|
116
|
+
('PO9',(-5.000, -3.000)),
|
|
117
|
+
('PO7',(-4.000, -3.000)),
|
|
118
|
+
('PO5',(-3.000, -3.000)),
|
|
119
|
+
('PO3',(-2.000, -3.000)),
|
|
120
|
+
('PO1',(-1.000, -3.000)),
|
|
121
|
+
('POz',(0.000, -3.000)),
|
|
122
|
+
('PO2',(1.000, -3.000)),
|
|
123
|
+
('PO4',(2.000, -3.000)),
|
|
124
|
+
('PO6',(3.000, -3.000)),
|
|
125
|
+
('PO8',(4.000, -3.000)),
|
|
126
|
+
('PO10',(5.000, -3.000)),
|
|
127
|
+
('POO9h',(-4.500, -3.250)),
|
|
128
|
+
('POO3h',(-2.000, -3.250)),
|
|
129
|
+
('POO4h',(2.000, -3.250)),
|
|
130
|
+
('POO10h',(4.500, -3.250)),
|
|
131
|
+
('O1',(-2.500, -3.750)),
|
|
132
|
+
('Oz',(0.000, -3.750)),
|
|
133
|
+
('O2',(2.500, -3.750)),
|
|
134
|
+
('OI1h',(1.500, -4.250)),
|
|
135
|
+
('OI2h',(-1.500, -4.250)),
|
|
136
|
+
('I1',(1.000, -4.500)),
|
|
137
|
+
('Iz',(0.000, -4.500)),
|
|
138
|
+
('I2',(-1.000, -4.500)),
|
|
139
|
+
)
|
|
140
|
+
|
|
141
|
+
|
|
142
|
+
def get_channelpos(channame, chan_pos_list):
|
|
143
|
+
if chan_pos_list[0] == 'angle':
|
|
144
|
+
return get_channelpos_from_angle(channame, chan_pos_list[1:])
|
|
145
|
+
elif chan_pos_list[0] == 'cartesian':
|
|
146
|
+
channame = channame.lower()
|
|
147
|
+
for name, coords in chan_pos_list[1:]:
|
|
148
|
+
if name.lower() == channame:
|
|
149
|
+
return coords[0], coords[1]
|
|
150
|
+
return None
|
|
151
|
+
else:
|
|
152
|
+
raise ValueError("Unknown first element "
|
|
153
|
+
"{:s} (should be type of positions)".format(
|
|
154
|
+
chan_pos_list[0]))
|
|
155
|
+
|
|
156
|
+
|
|
157
|
+
def get_channelpos_from_angle(channame, chan_pos_list=CHANNEL_10_20_APPROX):
|
|
158
|
+
"""Return the x/y position of a channel.
|
|
159
|
+
|
|
160
|
+
This method calculates the stereographic projection of a channel
|
|
161
|
+
from ``CHANNEL_10_20``, suitable for a scalp plot.
|
|
162
|
+
|
|
163
|
+
Parameters
|
|
164
|
+
----------
|
|
165
|
+
channame : str
|
|
166
|
+
Name of the channel, the search is case insensitive.
|
|
167
|
+
|
|
168
|
+
chan_pos_list=CHANNEL_10_20_APPROX,
|
|
169
|
+
interpolation='bilinear'
|
|
170
|
+
|
|
171
|
+
Returns
|
|
172
|
+
-------
|
|
173
|
+
x, y : float or None
|
|
174
|
+
The projected point on the plane if the point is known,
|
|
175
|
+
otherwise ``None``
|
|
176
|
+
|
|
177
|
+
Examples
|
|
178
|
+
--------
|
|
179
|
+
|
|
180
|
+
>>> plot.get_channelpos_from_angle('C2')
|
|
181
|
+
(0.1720792096741632, 0.0)
|
|
182
|
+
>>> # the channels are case insensitive
|
|
183
|
+
>>> plot.get_channelpos_from_angle('c2')
|
|
184
|
+
(0.1720792096741632, 0.0)
|
|
185
|
+
>>> # lookup for an invalid channel
|
|
186
|
+
>>> plot.get_channelpos_from_angle('foo')
|
|
187
|
+
None
|
|
188
|
+
|
|
189
|
+
"""
|
|
190
|
+
channame = channame.lower()
|
|
191
|
+
for i in chan_pos_list:
|
|
192
|
+
if i[0].lower() == channame:
|
|
193
|
+
# convert the 90/4th angular position into x, y, z
|
|
194
|
+
p = i[1]
|
|
195
|
+
ea, eb = p[0] * (90 / 4), p[1] * (90 / 4)
|
|
196
|
+
ea = ea * math.pi / 180
|
|
197
|
+
eb = eb * math.pi / 180
|
|
198
|
+
x = math.sin(ea) * math.cos(eb)
|
|
199
|
+
y = math.sin(eb)
|
|
200
|
+
z = math.cos(ea) * math.cos(eb)
|
|
201
|
+
# Calculate the stereographic projection.
|
|
202
|
+
# Given a unit sphere with radius ``r = 1`` and center at
|
|
203
|
+
# the origin. Project the point ``p = (x, y, z)`` from the
|
|
204
|
+
# sphere's South pole (0, 0, -1) on a plane on the sphere's
|
|
205
|
+
# North pole (0, 0, 1).
|
|
206
|
+
#
|
|
207
|
+
# The formula is:
|
|
208
|
+
#
|
|
209
|
+
# P' = P * (2r / (r + z))
|
|
210
|
+
#
|
|
211
|
+
# We changed the values to move the point of projection
|
|
212
|
+
# further below the south pole
|
|
213
|
+
mu = 1 / (1.3 + z)
|
|
214
|
+
x *= mu
|
|
215
|
+
y *= mu
|
|
216
|
+
return x, y
|
|
217
|
+
return None
|
|
@@ -3,6 +3,101 @@ import errno
|
|
|
3
3
|
import numpy as np
|
|
4
4
|
|
|
5
5
|
|
|
6
|
+
def corr(a, b):
|
|
7
|
+
"""
|
|
8
|
+
Computes correlation only between terms of a and terms of b, not within
|
|
9
|
+
a and b.
|
|
10
|
+
|
|
11
|
+
Parameters
|
|
12
|
+
----------
|
|
13
|
+
a, b: 2darray, features x samples
|
|
14
|
+
|
|
15
|
+
Returns
|
|
16
|
+
-------
|
|
17
|
+
Correlation between features in x and features in y
|
|
18
|
+
"""
|
|
19
|
+
# Difference to numpy:
|
|
20
|
+
# Correlation only between terms of x and y
|
|
21
|
+
# not between x and x or y and y
|
|
22
|
+
this_cov = cov(a, b)
|
|
23
|
+
return _cov_to_corr(this_cov, a, b)
|
|
24
|
+
|
|
25
|
+
|
|
26
|
+
def cov(a, b):
|
|
27
|
+
"""
|
|
28
|
+
Computes covariance only between terms of a and terms of b, not within
|
|
29
|
+
a and b.
|
|
30
|
+
|
|
31
|
+
Parameters
|
|
32
|
+
----------
|
|
33
|
+
a, b: 2darray, features x samples
|
|
34
|
+
|
|
35
|
+
Returns
|
|
36
|
+
-------
|
|
37
|
+
Covariance between features in x and features in y
|
|
38
|
+
"""
|
|
39
|
+
demeaned_a = a - np.mean(a, axis=1, keepdims=True)
|
|
40
|
+
demeaned_b = b - np.mean(b, axis=1, keepdims=True)
|
|
41
|
+
this_cov = np.dot(demeaned_a, demeaned_b.T) / (b.shape[1] - 1)
|
|
42
|
+
return this_cov
|
|
43
|
+
|
|
44
|
+
|
|
45
|
+
def _cov_to_corr(this_cov, a, b):
|
|
46
|
+
# computing "unbiased" corr
|
|
47
|
+
# ddof=1 for unbiased..
|
|
48
|
+
var_a = np.var(a, axis=1, ddof=1)
|
|
49
|
+
var_b = np.var(b, axis=1, ddof=1)
|
|
50
|
+
return _cov_and_var_to_corr(this_cov, var_a, var_b)
|
|
51
|
+
|
|
52
|
+
|
|
53
|
+
def _cov_and_var_to_corr(this_cov, var_a, var_b):
|
|
54
|
+
divisor = np.outer(np.sqrt(var_a), np.sqrt(var_b))
|
|
55
|
+
return this_cov / divisor
|
|
56
|
+
|
|
57
|
+
|
|
58
|
+
def wrap_reshape_apply_fn(stat_fn, a, b, axis_a, axis_b):
|
|
59
|
+
"""
|
|
60
|
+
Reshape two nd-arrays into 2d-arrays, apply function and reshape
|
|
61
|
+
result back.
|
|
62
|
+
Parameters
|
|
63
|
+
----------
|
|
64
|
+
stat_fn: function
|
|
65
|
+
Function to apply to 2d-arrays
|
|
66
|
+
a: nd-array: nd-array
|
|
67
|
+
b: nd-array
|
|
68
|
+
axis_a: int or list of int
|
|
69
|
+
sample axis
|
|
70
|
+
axis_b: int or list of int
|
|
71
|
+
sample axis
|
|
72
|
+
|
|
73
|
+
Returns
|
|
74
|
+
-------
|
|
75
|
+
result: nd-array
|
|
76
|
+
The result reshaped to remaining_dims_a + remaining_dims_b
|
|
77
|
+
"""
|
|
78
|
+
if not hasattr(axis_a, '__len__'):
|
|
79
|
+
axis_a = [axis_a]
|
|
80
|
+
if not hasattr(axis_b, '__len__'):
|
|
81
|
+
axis_b = [axis_b]
|
|
82
|
+
other_axis_a = [i for i in range(a.ndim) if i not in axis_a]
|
|
83
|
+
other_axis_b = [i for i in range(b.ndim) if i not in axis_b]
|
|
84
|
+
transposed_topo_a = a.transpose(tuple(other_axis_a) + tuple(axis_a))
|
|
85
|
+
n_stat_axis_a = [a.shape[i] for i in axis_a]
|
|
86
|
+
n_other_axis_a = [a.shape[i] for i in other_axis_a]
|
|
87
|
+
flat_topo_a = transposed_topo_a.reshape(np.prod(n_other_axis_a),
|
|
88
|
+
np.prod(n_stat_axis_a))
|
|
89
|
+
transposed_topo_b = b.transpose(tuple(other_axis_b) + tuple(axis_b))
|
|
90
|
+
n_stat_axis_b = [b.shape[i] for i in axis_b]
|
|
91
|
+
n_other_axis_b = [b.shape[i] for i in other_axis_b]
|
|
92
|
+
flat_topo_b = transposed_topo_b.reshape(np.prod(n_other_axis_b),
|
|
93
|
+
np.prod(n_stat_axis_b))
|
|
94
|
+
assert np.array_equal(n_stat_axis_a, n_stat_axis_b)
|
|
95
|
+
stat_result = stat_fn(flat_topo_a, flat_topo_b)
|
|
96
|
+
topo_result = stat_result.reshape(
|
|
97
|
+
tuple(n_other_axis_a) + tuple(n_other_axis_b))
|
|
98
|
+
return topo_result
|
|
99
|
+
|
|
100
|
+
|
|
6
101
|
class FuncAndArgs(object):
|
|
7
102
|
"""Container for a function and its arguments.
|
|
8
103
|
Useful in case you want to pass a function and its arguments
|
|
@@ -108,9 +108,9 @@ author = 'Robin Tibor Schirrmeister'
|
|
|
108
108
|
# built documents.
|
|
109
109
|
#
|
|
110
110
|
# The short X.Y version.
|
|
111
|
-
version = '0.1.
|
|
111
|
+
version = '0.1.6'
|
|
112
112
|
# The full version, including alpha/beta/rc tags.
|
|
113
|
-
release = '0.1.
|
|
113
|
+
release = '0.1.6'
|
|
114
114
|
|
|
115
115
|
# The language for content autogenerated by Sphinx. Refer to documentation
|
|
116
116
|
# for a list of supported languages.
|
|
@@ -418,10 +418,10 @@
|
|
|
418
418
|
" for batch_X, batch_y in iterator.get_batches(train_set, shuffle=False):\n",
|
|
419
419
|
" net_in = np_to_var(batch_X)\n",
|
|
420
420
|
" if cuda:\n",
|
|
421
|
-
" net_in.cuda()\n",
|
|
421
|
+
" net_in = net_in.cuda()\n",
|
|
422
422
|
" net_target = np_to_var(batch_y)\n",
|
|
423
423
|
" if cuda:\n",
|
|
424
|
-
" net_target.cuda()\n",
|
|
424
|
+
" net_target = net_target.cuda()\n",
|
|
425
425
|
" # Remove gradients of last backward pass from all parameters \n",
|
|
426
426
|
" optimizer.zero_grad()\n",
|
|
427
427
|
" outputs = model(net_in)\n",
|
|
@@ -445,10 +445,10 @@
|
|
|
445
445
|
" for batch_X, batch_y in iterator.get_batches(dataset, shuffle=False):\n",
|
|
446
446
|
" net_in = np_to_var(batch_X)\n",
|
|
447
447
|
" if cuda:\n",
|
|
448
|
-
" net_in.cuda()\n",
|
|
448
|
+
" net_in = net_in.cuda()\n",
|
|
449
449
|
" net_target = np_to_var(batch_y)\n",
|
|
450
450
|
" if cuda:\n",
|
|
451
|
-
" net_target.cuda()\n",
|
|
451
|
+
" net_target = net_target.cuda()\n",
|
|
452
452
|
" outputs = model(net_in)\n",
|
|
453
453
|
" all_preds.append(var_to_np(outputs))\n",
|
|
454
454
|
" outputs = th.mean(outputs, dim=2)[:,:,0]\n",
|
|
@@ -464,10 +464,10 @@
|
|
|
464
464
|
" batch_y = train_set.y[i_trials]\n",
|
|
465
465
|
" net_in = np_to_var(batch_X)\n",
|
|
466
466
|
" if cuda:\n",
|
|
467
|
-
" net_in.cuda()\n",
|
|
467
|
+
" net_in = net_in.cuda()\n",
|
|
468
468
|
" net_target = np_to_var(batch_y)\n",
|
|
469
469
|
" if cuda:\n",
|
|
470
|
-
" net_target.cuda()\n",
|
|
470
|
+
" net_target = net_target.cuda()\n",
|
|
471
471
|
" # Remove gradients of last backward pass from all parameters \n",
|
|
472
472
|
" optimizer.zero_grad()\n",
|
|
473
473
|
" # Compute outputs of the network\n",
|
|
@@ -487,10 +487,10 @@
|
|
|
487
487
|
" # for such smaller datasets. Otherwise we would have to use batches.\n",
|
|
488
488
|
" net_in = np_to_var(dataset.X[:,:,:,None])\n",
|
|
489
489
|
" if cuda:\n",
|
|
490
|
-
" net_in.cuda()\n",
|
|
490
|
+
" net_in = net_in.cuda()\n",
|
|
491
491
|
" net_target = np_to_var(dataset.y)\n",
|
|
492
492
|
" if cuda:\n",
|
|
493
|
-
" net_target.cuda()\n",
|
|
493
|
+
" net_target = net_target.cuda()\n",
|
|
494
494
|
" outputs = model(net_in)\n",
|
|
495
495
|
" loss = F.nll_loss(outputs, net_target)\n",
|
|
496
496
|
" print(\"{:6s} Loss: {:.5f}\".format(\n",
|
|
@@ -536,6 +536,11 @@
|
|
|
536
536
|
},
|
|
537
537
|
"outputs": [],
|
|
538
538
|
"source": [
|
|
539
|
+
"import mne\n",
|
|
540
|
+
"import numpy as np\n",
|
|
541
|
+
"from mne.io import concatenate_raws\n",
|
|
542
|
+
"from braindecode.datautil.signal_target import SignalAndTarget\n",
|
|
543
|
+
"\n",
|
|
539
544
|
"physionet_paths = [ mne.datasets.eegbci.load_data(sub_id,[4,8,12,]) for sub_id in range(1,51)]\n",
|
|
540
545
|
"physionet_paths = np.concatenate(physionet_paths)\n",
|
|
541
546
|
"parts = [mne.io.read_raw_edf(path, preload=True,stim_channel='auto')\n",
|