MRG Ica plot #2251

jaeilepp · 2015-06-30T12:35:58Z

Here's a plotter for ICA components (item at #2213). The interactive selection doesn't work yet and neither does scrolling horizontally, but I wanted to show what I've done so far, because I'm not sure if this is what you guys wanted in the first place. I'm no expert in using ICA, so comments are welcome.

The plotter has basically the same interactive functions as does other browser functions. This PR also includes quite a bit of re-factoring of the existing browser functions.

larsoner · 2015-06-30T12:45:13Z

I was going to suggest that you do some heavy refactoring to unify bits of code but it sounds like you're already on it! Excellent.

dengemann · 2015-06-30T12:48:21Z

mne/preprocessing/ica.py

we should avoid mutable defaults e.g. lists.

agramfort · 2015-06-30T14:17:49Z

cool

let us know when it's ready. If not yet.

mainakjas · 2015-06-30T16:28:15Z

awesome ! how should I test this? Is there an example I can run?

jaeilepp · 2015-07-01T10:02:10Z

I added a line in the example that uses this function. If you're gonna run it many times, it's probably wise to save the ica solution and use that.

jaeilepp · 2015-07-02T10:03:51Z

I made it so that the plotting function now returns the updated list of components marked for exclusion, as selected by the user. Would it be okay to exclude the components straight away on closing the figure (ica.exclude)? At the moment it only updates the list, but is inconsistent with other plot functions because it returns a tuple of figure and list instead of just the figure.

dengemann · 2015-07-02T15:06:20Z

mne/preprocessing/ica.py

did we already discuss why not simply updating ICA.exclude?

No

On 2 juil. 2015, at 16:06, Denis A. Engemann notifications@github.com wrote:

In mne/preprocessing/ica.py:

Defaults to (1., 0., 0.) (red).

show : bool

Show figures if True. Defaults to True.

block : bool

Whether to halt program execution until the figure is closed.

Useful for selecting components for exclusion on the fly

(click on line). May not work on all systems / platforms.

Defaults to False.

Returns

fig : Instance of matplotlib.figure.Figure

The figure.

exclude : list

Updated list of components marked for exclusion.

"""
did we already discuss why not simply updating ICA.exclude?

—
Reply to this email directly or view it on GitHub.

dengemann · 2015-07-02T15:07:07Z

Would it be okay to exclude the components straight away on closing the figure (ica.exclude)?

I would opt for that.

agramfort · 2015-07-02T15:20:15Z

+1

On 2 juil. 2015, at 16:07, Denis A. Engemann notifications@github.com wrote:

Would it be okay to exclude the components straight away on closing the figure (ica.exclude)?

I would opt for that.

—
Reply to this email directly or view it on GitHub.

mainakjas · 2015-07-02T19:25:04Z

examples/preprocessing/plot_ica_from_raw.py

This example is quite crazy. It opens 12 or 13 figures ! anything can be done about it? cc @dengemann

I would like to leave it like this, it has been very useful for tracking bugs in the past.

mainakjas · 2015-07-02T19:26:08Z

There is no help button here. I suppose you'll add that soon?

jaeilepp · 2015-07-03T08:16:08Z

The help button positioning is partly dependent on the layout remake at #2214. I'd rather wait for it to be merged before implementing it here.

dengemann · 2015-07-03T08:18:54Z

@jaeilepp @mainakjas @agramfort I still think we should just overwrite plot_sources or so. Not sure we should add new methods.

mainakjas · 2015-07-03T08:23:04Z

On Fri, Jul 3, 2015 at 10:18 AM, Denis A. Engemann <notifications@github.com

wrote:

@jaeilepp https://github.com/jaeilepp @mainakjas
https://github.com/mainakjas @agramfort https://github.com/agramfort
I still think we should just overwrite plot_sources or so. Not sure we
should add new methods.

+1 I agree. What we have is a more sophisticated version of plot_sources,
right?

—
Reply to this email directly or view it on GitHub
#2251 (comment)
.

dengemann · 2015-07-03T08:24:02Z

plot_sources,
right?

yes and plot_sources was the weakest of the plotting functions at least for raw. The evoked counterpart is really useful.

dengemann · 2015-07-03T08:24:46Z

... which reminds us of the fact that we should also recycle the epochs viz here. I would suspect that is at least as useful as the raw viz for ICA.

agramfort · 2015-07-03T12:31:02Z

+1 for recycling raw and epochs browser here.

jaeilepp · 2015-07-06T10:34:02Z

Epoch plotter implemented. Is there any sense in marking bad epochs here? Should it be disabled? Also, is the exclude parameter needed? Ica object already has an exclude attribute, which is basically the same. Same goes for raw plotter. So basically the use case would change so that the user updates the ica.exclude before plotting to see the components marked for exclusion.

dengemann · 2015-07-06T10:45:43Z

No sense, so yes, it should be disabled

On 06 Jul 2015, at 12:34, jaeilepp notifications@github.com wrote:

Epoch plotter implemented. Is there any sense in marking bad epochs here? Should it be disabled? Also, is the exclude parameter needed? Ica object already has an exclude attribute, which is basically the same. Same goes for raw plotter. So basically the use case would change so that the user updates the ica.exclude before plotting to see the components marked for exclusion.

—
Reply to this email directly or view it on GitHub.

jaeilepp · 2015-07-06T12:59:07Z

I changed the plot_sources to use these new functions for epochs and raw instances. See if you're happy.

agramfort · 2015-07-06T13:10:06Z

doc/source/whats_new.rst

agramfort · 2015-07-06T13:19:27Z

works great for me did not review full code though. Usability feedback. Not for this PR but it would be great to preview the topography of source from the browser. I see some nice alpha waves on raw examples and it's frustrating not to be able to see the associated topography.

jaeilepp · 2015-07-07T10:25:41Z

Usability feedback. Not for this PR but it would be great to preview the topography of source from the browser.

Did you mean like this? I implemented it here, since it was quite easy to do.

The component can be shown by clicking on the y axis label. I don't think it's going to work on every data since it relies on getting the layout from the data.
The indexing is actually inconsistent (ICA 109 = IC 108). I'll change that...

dengemann · 2015-07-07T10:37:12Z

The indexing is actually inconsistent (ICA 109 = IC 108). I'll change that...

Now it's a good moment to attack that (cc @teonlamont). I'd do it at the plotting level, not at the level of the sources as raw function.

I'm still wondering though what's better. It's actually correct: ICA component #109 has the python index 108, wheras component #0 does not exist. We do the same for the channels btw.

jaeilepp · 2015-07-07T11:20:32Z

I'm still wondering though what's better. It's actually correct: ICA component #109 has the python index 108, wheras component #0 does not exist. We do the same for the channels btw.

Yeah, I intuitively thought that the indexing should start from 1, but all the existing plotting functions seem to be using 0 for ICA.

jaeilepp · 2015-07-07T12:21:30Z

I guess this is ready for review, but it still needs some testing.

teonbrooks · 2015-07-07T13:30:12Z

closes #2000 #2043.
are we able to mark bad ICs from the browser like in the raw/epochs viewer?

also, what's the verdict on the indexing, are we plotting the ICs to start at 1 e.g. ICA 001 for all the plots now? do we still exclude based on the python index? if so, should we support exclude to take IC label to be in line with info['bads']? I think @agramfort and me mentioned it as a solution at one point #2019 (comment). for me, as long as we have a consistent use of names across the board, I'm happy.

btw, the topo view is super useful

…ually from example.

jaeilepp · 2015-07-14T12:02:01Z

Ready for review.

agramfort · 2015-07-14T12:59:22Z

I am happy

+1 for merge

@dengemann ?

dengemann · 2015-07-14T13:30:38Z

@jaeilepp we still don't see the EOG channels / ECG channels in the ICA time series, do we? I'll merge but please open an issue for later. Thanks for the wonderful work!

MRG Ica plot

agramfort · 2015-07-14T13:34:16Z

great !

dengemann reviewed Jun 30, 2015
View reviewed changes

mne/preprocessing/ica.py Outdated

Copy link

Member

dengemann Jun 30, 2015

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

we should avoid mutable defaults e.g. lists.

dengemann reviewed Jul 2, 2015
View reviewed changes

mainakjas reviewed Jul 2, 2015
View reviewed changes

jaeilepp force-pushed the ica-plot branch from d070ad3 to a0d601e Compare July 6, 2015 12:50

agramfort reviewed Jul 6, 2015
View reviewed changes

doc/source/whats_new.rst Outdated

Copy link

Member

agramfort Jul 6, 2015

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

bad indent

jaeilepp force-pushed the ica-plot branch from a0d601e to 0e6c9b1 Compare July 7, 2015 08:04

jaeilepp changed the title ~~WIP Ica plot~~ Ica plot Jul 7, 2015

jaeilepp and others added 20 commits July 14, 2015 12:23

Updated examples and whats_new. Docstrings.

51b5f00

Fix to whats_new. Tests.

a075233

Fixes. Components plotted on click to label.

1a5752c

Fixes.

53e8239

Bad channel selection by clicking label in mne_browse_raw.

c83336a

Separate topomaps for eeg, mag and grad. Fixes.

e3a5b65

Cleaning. Docs.

063a221

Docs. Tests. Checks.

fff6c7f

Customized help dialog. Fixes.

d6d33fa

Help box for browse_raw.

265992a

Added interactive mode for viewing ica topos.

a3781a1

Tests.

3682087

Reorganizing. Fix.

9fc33c6

Unselecting removes duplicate bads. Removed adding of bad channel man…

5e817d9

…ually from example.

Component name format. Manual bad channel back to example.

696f369

Fix to exclusion.

71daaeb

Fix.

ec50bd4

fixes fixes yeah

f8d234f

Fix to raw sources.

95e0ac1

Fix to channels per view.

00f65f5

Small fixes.

57cd76b

jaeilepp changed the title ~~Ica plot~~ MRG Ica plot Jul 14, 2015

dengemann added a commit that referenced this pull request Jul 14, 2015

Merge pull request #2251 from jaeilepp/ica-plot

ce320f5

MRG Ica plot

dengemann merged commit ce320f5 into mne-tools:master Jul 14, 2015

jaeilepp mentioned this pull request Jul 15, 2015

[MRG] EOG and ECG channels added to the bottom of ICA plot. #2303

Merged

teonbrooks mentioned this pull request Jul 24, 2015

ICA labeling #2043

Closed

Uh oh!

MRG Ica plot #2251

MRG Ica plot #2251

Uh oh!

Conversation

jaeilepp commented Jun 30, 2015

Uh oh!

larsoner commented Jun 30, 2015 via email

Uh oh!

dengemann Jun 30, 2015

Choose a reason for hiding this comment

Uh oh!

agramfort commented Jun 30, 2015

Uh oh!

mainakjas commented Jun 30, 2015

Uh oh!

jaeilepp commented Jul 1, 2015

Uh oh!

jaeilepp commented Jul 2, 2015

Uh oh!

dengemann Jul 2, 2015

Choose a reason for hiding this comment

Uh oh!

agramfort Jul 2, 2015

Choose a reason for hiding this comment

Uh oh!

dengemann commented Jul 2, 2015

Uh oh!

agramfort commented Jul 2, 2015

Uh oh!

mainakjas Jul 2, 2015

Choose a reason for hiding this comment

Uh oh!

dengemann Jul 3, 2015

Choose a reason for hiding this comment

Uh oh!

mainakjas commented Jul 2, 2015

Uh oh!

jaeilepp commented Jul 3, 2015

Uh oh!

dengemann commented Jul 3, 2015

Uh oh!

mainakjas commented Jul 3, 2015

Uh oh!

dengemann commented Jul 3, 2015

Uh oh!

dengemann commented Jul 3, 2015

Uh oh!

agramfort commented Jul 3, 2015 via email

Uh oh!

jaeilepp commented Jul 6, 2015

Uh oh!

dengemann commented Jul 6, 2015

Uh oh!

jaeilepp commented Jul 6, 2015

Uh oh!

agramfort Jul 6, 2015

Choose a reason for hiding this comment

Uh oh!

agramfort commented Jul 6, 2015 via email

Uh oh!

jaeilepp commented Jul 7, 2015

Uh oh!

dengemann commented Jul 7, 2015

Uh oh!

jaeilepp commented Jul 7, 2015

Uh oh!

jaeilepp commented Jul 7, 2015

Uh oh!

teonbrooks commented Jul 7, 2015

Uh oh!

jaeilepp commented Jul 14, 2015

Uh oh!

agramfort commented Jul 14, 2015

Uh oh!

dengemann commented Jul 14, 2015

Uh oh!

agramfort commented Jul 14, 2015 via email

Uh oh!

Reviewers