New datasets and parameter confirmation #32

vinay-jayaram · 2018-04-05T11:12:33Z

This PR includes data mentioned in 1 from #1 (checked items) as well as a review of the epoching parameters that were mentioned as defaults. In particular:

Many datasets had inconsistent start and end times when compared with the publications. I do see how cutting the beginning/end of trials eliminates ramp up/down effects, but it's not applied consistently (datasets like Alex MI use everything) and so I went through the publications and changed all epoching parameters to the cues described in the papers.
We need to standardize how epochs are given to the data -- I would suggest a function set_epoch that every dataset needs to implement, so that paradigms can go through and change the epoching if they want to.
Weibo 2014 and Zhou 2016 datasets are added and have been tested with a function in moabb.tests.datasets that just runs through all subjects and confirms they can be downloaded and loaded properly.

Many changes also detailed in pull request

1. Added task_interval to all datasets so we know how imagination time and epoching interval are related 2. Added docs to BaseDataset so it's less confusing 3. Re-added ImageryNClass since LvR is too restrictive for our current datasets 4. Made paradigms look at the selected_events and not event_id parameters since we may only care about e.g. motor imageries or cognitive imagery 5. Separated filter bank and single-filter paradigms

vinay-jayaram · 2018-04-05T13:31:09Z

I solved the epoching issue by simply giving each dataset a parameter that specifies when the imagery starts and stops with respect to the events, so the interval is always relative to imagery start and not event (since events are dataset-specific and sometimes have baselines and other things)

alexandrebarachant · 2018-04-05T13:45:09Z

Many datasets had inconsistent start and end times when compared with the publications. I do see how cutting the beginning/end of trials eliminates ramp up/down effects, but it's not applied consistently (datasets like Alex MI use everything) and so I went through the publications and changed all epoching parameters to the cues described in the papers.

Good call on this one. my initial intention was to skip some of the ERP related to the cue, but i have been sloppy. better to put the true events timing.

We need to standardize how epochs are given to the data -- I would suggest a function set_epoch that every dataset needs to implement, so that paradigms can go through and change the epoching if they want to.

We could do that using the interval from the paradigm. in that case the interval is relative to the timing of the dataset. for example

interval = [1, None] will epoch 1 second after the event, until the end.
interval = [1, 3] will take from 1 to 3 second
interval = [1, -1] from one to -1 (not sure about this one)

to implement this, we can use the function epochs.crop() from mne.

Weibo 2014 and Zhou 2016 datasets are added and have been tested with a function in moabb.tests.datasets that just runs through all subjects and confirms they can be downloaded and loaded properly.

We have this in tests,.download.py i put this out of the regular test suit to avoid travis downloading a bunch of data every time. I yet have to find the solution to skip those test on travis but run them localy.

I'n going to review your PR in detail, just tell me when you are done.

alexandrebarachant · 2018-04-05T13:50:58Z

moabb/datasets/base.py

        if not isinstance(subjects, list):
            raise(ValueError("subjects must be a list"))

        self.subject_list = subjects
        self.n_sessions = sessions_per_subject
        self.event_id = events
+        self.selected_events = events.copy()


I dont think this is a good way to proceed. The paradigm should not modify the dataset object.
This expose to corner case where the same dataset object is used with different paradigm

hm. I will think of another solution

alexandrebarachant · 2018-04-05T13:53:22Z

moabb/paradigms/motor_imagery.py

@@ -69,21 +69,22 @@ def process_raw(self, raw, dataset):

        # get event id
        if self.events is None:
-            event_id = dataset.event_id
+            event_id = dataset.selected_events


See my comment. If we restrict to a certain set of events, this should be done by the paradigm. the dataset is context independent.

alexandrebarachant · 2018-04-05T13:54:10Z

moabb/paradigms/motor_imagery.py

            # skip raw if no event found
            return

        # get interval
        if self.interval is None:
            tmin, tmax = dataset.interval
        else:
+            # TODO: make this work with varying offsets since not all data starts at 0


lets use epochs.crop() after epoching

what if someone wants to do a baseline though? Wouldn't it be better to define the window while making the Epochs object, instead of making it with the full task interval and then cropping it after?

alexandrebarachant · 2018-04-05T13:55:42Z

moabb/datasets/Weibo2014.py

+FILES.append('https://dataverse.harvard.edu/api/access/datafile/2499179')
+
+
+def eeg_data_path(base_path, subject):


Can we somehow avoid this code duplication ?

alexandrebarachant · 2018-04-05T13:58:13Z

moabb/datasets/Weibo2014.py

+
+
+class Weibo2014(BaseDataset):
+    """Weibo 2014 Motor Imagery dataset"""


This would be good if we can start documenting what contain each dataset.

We can copy / paste and modify the dataset description from the publication.

We should also add a Reference section where we cite the paper, so it appears in the doc. for example : https://github.com/scikit-learn/scikit-learn/blob/a24c8b46/sklearn/ensemble/weight_boosting.py#L364

alexandrebarachant · 2018-04-05T13:58:57Z

moabb/datasets/Zhou2016.py

+
+
+class Zhou2016(BaseDataset):
+    """Zhou 2016 Imagery dataset"""


idem, lets document this properly

alexandrebarachant · 2018-04-05T13:59:46Z

moabb/datasets/Zhou2016.py

@@ -0,0 +1,85 @@
+'''


we dont really need a docstring module for this one. its not parsed by the doc.
you can still let this doc here, but put everything you want to see on the doc in the class docstring

alexandrebarachant · 2018-04-05T14:01:20Z

moabb/datasets/__init__.py

@@ -11,3 +11,5 @@
 from .openvibe_mi import OpenvibeMI
 from .bbci_eeg_fnirs import BBCIEEGfNIRS
 from .upper_limb import UpperLimb
+from .Weibo2014 import Weibo2014


please, add them in docs/source/datasets.rst

alexandrebarachant · 2018-04-05T14:04:55Z

moabb/datasets/base.py

+            Number of sessions per subject
+
+        events: dict of string: int
+            String codes for events matched with labels in the stim channel. Currently imagery codes codes can include:


we also have elbow and other stuff.

I'm wondering if we should start using MNE hierarchical event definition.

for example, you can define an event as hand/left and hand/right, which allow to select all hand event after epoching by doing Epochs['hand']. But this is another discussion

Well we lose nothing by adding it so why not, I'll go through and change all left_hand to hand/left etc --although there is one more level to worry about, of imagined vs actual

you can do imagined/hand/left, etc

but we can skip this for now, and see how we can deal with that later.

alexandrebarachant · 2018-04-05T14:16:49Z

moabb/datasets/base.py

        self.code = code
        self.interval = interval
+        if task_interval is None:


I'm still wondering if this is the right way to proceed.
We can set this to be the default interval, i.e. interval from the start to the end of MI.

This type of thing is very specific to MI.

I think it generalizes OK, the case of ERPs is just that task_interval[0] is 0 and task_interval[1] is the length of time until the next trigger

alexandrebarachant · 2018-04-05T14:17:53Z

moabb/paradigms/motor_imagery.py

@@ -125,8 +126,20 @@ def scoring(self):
        return 'accuracy'


+class MotorImagery(BaseMotorImagery):
+
+    def __init__(self, bp_low=8, bp_high=32, **kwargs):


can you use fmin and fmax, to be consistent with MNE.

1. Use new function used_events to choose events for a paradigm within the paradigm 2. Fixed bug in Physionet where hand and feet imagery couldn't be used together 3. Added documentation to new datasets and put them in .rst file 4. Simplified download code in Weibo2014

vinay-jayaram · 2018-04-05T15:13:11Z

all fixes except the changing of event codes implemented.

alexandrebarachant · 2018-04-05T17:17:07Z

moabb/paradigms/motor_imagery.py


        # pick events, based on event_id
        try:
            events = mne.pick_events(events, include=list(event_id.values()))
-        except RuntimeError:
+            # not all runs have all events. Reduce event dict to only occurring events
+            events_in_run = np.unique(events[:, 2])


I think this is not necessary if we set the parameter on_missing='ignore' in the mne.Epochs call

Will check this out

alexandrebarachant · 2018-04-05T17:22:19Z

moabb/datasets/Zhou2016.py

+    Abstract
+    ------------
+
+    Independent component analysis (ICA) as a promising spatial filtering method


Sorry, i did not express myself very clearly. I was thinking about the part of the article that describes the dataset.
The goal is to describe the typo of motor imagery, the timing, the task, etc.

For example, this part :

Throughout the experiment, the subject sat in a comfortable armchair facing a computer screen. The duration of each trial was 10 s as shown in Fig 1. A trial started by a short beep indicating 1 s preparation time, and followed by a red arrow pointing randomly to three directions (left, right, or bottom) lasting for 5 s and then presented a black screen for 4 s. The subject was instructed to immediately perform the imagination tasks of the left hand, right hand or foot movement respectively according to the cue direction, and try to relax during the black screen.

and

Every subject went through three sessions, each of which contained two consecutive runs with several minutes inter-run breaks, and each run comprised 75 trials (25 trials per class). The intervals between two sessions varied from several days to several months.

alexandrebarachant · 2018-04-05T17:27:33Z

moabb/datasets/Weibo2014.py

+
+class Weibo2014(BaseDataset):
+    """Weibo 2014 Motor Imagery dataset [1]
+


Can you do the same,

In order to investigate the differences of the EEG patterns as well as cognitive process between simple limb motor imagery and compound limb motor imagery, seven kinds of mental tasks have been designed, involving three tasks of simple limb motor imagery (left hand, right hand, feet), three tasks of compound limb motor imagery combining hand with hand/foot (both hands, left hand combined with right foot, right hand combined with left foot) and rest state.

At the beginning of each trial (8 seconds), a white circle appeared at the center of the monitor. After 2 seconds, a red circle (preparation cue) appeared for 1 second to remind the subjects of paying attention to the character indication next. Then red circle disappeared and character indication (‘Left Hand’, ‘Left Hand & Right Foot’, et al) was presented on the screen for 4 seconds, during which the participants were asked to perform kinesthetic motor imagery rather than a visual type of imagery while avoiding any muscle movement. After 7 seconds, ‘Rest’ was presented for 1 second before next trial (Fig. 1(a)). The experiments were divided into 9 sections, involving 8 sections consisting of 60 trials each for six kinds of MI tasks (10 trials for each MI task in one section) and one section consisting of 80 trials for rest state. The sequence of six MI tasks was randomized. Intersection break was about 5 to 10 minutes.

alexandrebarachant · 2018-04-05T17:30:20Z

moabb/datasets/Weibo2014.py

+                  update_path=None, verbose=None):
+        if subject not in self.subject_list:
+            raise(ValueError("Invalid subject number"))
+        key = 'MNE_DATASETS_WEIBO2014_PATH'


I usually got a weird warning about the non standard MNE key. Can we track down this kind of thing and change the keys for new datasets.

alexandrebarachant · 2018-04-05T17:31:26Z

moabb/datasets/Zhou2016.py

+                  update_path=None, verbose=None):
+        if subject not in self.subject_list:
+            raise(ValueError("Invalid subject number"))
+        key = 'MNE_DATASETS_ZHOU2016_PATH'


idem, lets change the key for MNE standard key (whatever the standard is)

NVM, this is a warning because it's not in the list a pre-approved config name. https://github.com/mne-tools/mne-python/blob/master/mne/utils.py#L1478

We should try and deal with this at some point though, as the list keeps growing...maybe our own config file?

alexandrebarachant · 2018-04-05T17:31:56Z

moabb/datasets/Zhou2016.py

+            raise(ValueError("Invalid subject number"))
+        key = 'MNE_DATASETS_ZHOU2016_PATH'
+        path = _get_path(path, key, "Zhou 2016")
+        _do_path_update(path, True, key, "Zhou 2016")


do we have any other option than forcing the path ?

this is what we do everywhere -- it should be another PR I think, revamping the download system

Yep we should. Let's keep this for another PR (not a priority)

alexandrebarachant · 2018-04-05T17:32:43Z

moabb/datasets/base.py

+            Number of sessions per subject
+
+        events: dict of string: int
+            String codes for events matched with labels in the stim channel. Currently imagery codes codes can include:


you can do imagined/hand/left, etc

alexandrebarachant · 2018-04-05T17:33:29Z

moabb/datasets/base.py

+            Number of sessions per subject
+
+        events: dict of string: int
+            String codes for events matched with labels in the stim channel. Currently imagery codes codes can include:


but we can skip this for now, and see how we can deal with that later.

alexandrebarachant · 2018-04-06T01:01:26Z

moabb/datasets/base.py

+        paradigm: ['p300','imagery']
+            Defines what sort of dataset this is (currently only imagery is implemented)
+
+        task_interval: list of 2 entries or None


Sorry for being picky, but this change make me inconfortable, and i'm trying to avoid complicating the API more than necessary.

I do not think we need to have a double definition of time interval. The time interval should correspond to the interval of the task i.e. the motor imagery. so i would actually replace the interval by your task task_interval.

The actual timing of the trial does not bring us anything.
We changed the API significantly, and it is the paradigm itself that is doing the epoching, so we don't event need to expose tmin and tmax in the dataset anymore. If the user want to play with cropping the epoch, it is more natural to do it at the level of the paradigm (exactly like events or channels).

To my point of view, datasets are immuable objects, they are here to abstract the data as it has been recorded.

does it make sense ?

yes, I think that does make the most sense. I'll change that everywhere

alexandrebarachant · 2018-04-06T05:25:42Z

moabb/datasets/bnci.py

@@ -642,49 +642,52 @@ def data_path(self, subject, path=None, force_update=False,
 class BNCI2014001(MNEBNCI):
    """BNCI 2014-001 Motor Imagery dataset"""

-    def __init__(self, tmin=3.5, tmax=5.5):
+    def __init__(self, tmin=0.5, tmax=4):


according to this : https://lampx.tugraz.at/~bci/database/001-2014/description.pdf
the motor imagery start at 3 second after the beginning of the trial (trigger).
I dont get you new your new parameters

i get a decrease in perf with this parameters

yep, the beep is at the trigger , the cue at 2 s, and the end around 6 s (epoch average at based on MNE epoching from the events)

import mne from moabb.datasets import * from moabb.paradigms import LeftRightImagery dataset = BNCI2014001() subject = 2 data = dataset.get_data([subject]) raw = mne.concatenate_raws(list(data[subject]['session_T'].values())) raw.filter(1, 20) events = mne.find_events(raw) raw.pick_types(eeg=True) events_id = dataset.event_id epochs = mne.Epochs(raw, events, event_id=events_id, tmin=-2, tmax=8, baseline=None, preload=True) epochs.average().plot()

Sorry--you're probably right, I was confused because I didn't know if the trigger was at the cue or the beep (I think in the eeg/fnirs one it's the cue...)

Needs to be prettied up at some point...

- use on_missing - intervals specified by paradigm are relative to imagery start - added min interval search to utils - paradigm datasets property now use interval search

vinay-jayaram · 2018-04-06T11:32:09Z

Making the BaseMotorImagery class abstract breaks the paradigm exploration script -- but I think it's reasonable to make it so, since just "MotorImagery" isn't specific enough for a paradigm, since it's not obvious how events should be handled. I'd say we switch to FilterBankImageryNClass and ImageryNClass

alexandrebarachant · 2018-04-06T13:02:37Z

moabb/datasets/bbci_eeg_fnirs.py

+
+    Data Acquisition
+    ----------------------------------------
+    EEG and NIRS data was collected in an ordinary bright room. EEG data was recorded by a multichannel BrainAmp EEG amplifier with thirty active electrodes (Brain Products GmbH, Gilching, Germany) with linked mastoids reference at 1000 Hz sampling rate. The EEG amplifier was also used to measure the electrooculogram (EOG), electrocardiogram (ECG) and respiration with a piezo based breathing belt. Thirty EEG electrodes were placed on a custom-made stretchy fabric cap (EASYCAP GmbH, Herrsching am Ammersee, Germany) and placed according to the international 10-5 system (AFp1, AFp2, AFF1h, AFF2h, AFF5h, AFF6h, F3, F4, F7, F8, FCC3h, FCC4h, FCC5h, FCC6h, T7, T8, Cz, CCP3h, CCP4h, CCP5h, CCP6h, Pz, P3, P4, P7, P8, PPO1h, PPO2h, POO1, POO2 and Fz for ground electrode).


FYI, Try to limit the line length at 79 char.
I'm soon gonna activate flake8 on travis, to enforce coding style, and this will make the test crash

alexandrebarachant · 2018-04-06T13:08:21Z

moabb/datasets/bnci.py

+    second 7.5 the screen went blank and a random interval between 1.0 and 2.0
+    seconds was added to the trial.
+
+    [1] R. Leeb, F. Lee, C. Keinrath, R. Scherer, H. Bischof, G. Pfurtscheller.


chech the docstring syntax for what i did in Zhou2016. if you do it right, it create a link when you reference [1]_

BTW, thanks for adding those docstring !

you can look how it render by generating the doc (cd docs & make html). the output files are in docs/build/html

alexandrebarachant · 2018-04-06T13:13:43Z

Making the BaseMotorImagery class abstract breaks the paradigm exploration script -- but I think it's reasonable to make it so, since just "MotorImagery" isn't specific enough for a paradigm, since it's not obvious how events should be handled. I'd say we switch to FilterBankImageryNClass and ImageryNClass

I'm thinking that we could just rename ImageryNClass to MotorImagery, and use it as our main most generic class. We could make something that use all events if n_classes is None, which will cover all the usecases.

vinay-jayaram · 2018-04-06T14:06:58Z

The number of classes varies so much though--I'm not sure it would be a valid comparison (and it would ruin the statistics) if we let them have different numbers. I'd force them to provide n_classes (or at least have the default be something as it is now), but keep it the same otherwise. In the future we can add logic that restricts the classes to be motor imagery but as of now that would be a huge pain

also fixed bug in ImageryNClass init

BaseMotorImagery is still abstract

alexandrebarachant · 2018-04-07T16:41:32Z

moabb/paradigms/motor_imagery.py

@@ -58,6 +60,10 @@ def verify(self, dataset):

        # we should verify list of channels, somehow

+    @abc.abstractmethod
+    def used_events(self, dataset):


The name is not super clear, i would suggest to rename this function to get_event_id()
but it would be helpful to document this abstract method so we know what we should implement in child class.

yeah if you have a better name go ahead and change it, I didn't put too much thought into that. paradigm_specific_events() maybe?

i would call it get_events() or something like that.

sounds good 👍

alexandrebarachant

Thanks a lot for this one. This is really really good :)

Just one addition, can you make sure the doc in docs/sources/paradigms.rst is up to date with the names of the paradigm classes.

You can merge after that.

As a side note, i have been more picky than usual on the PR review. I know adding documentation and other non code related stuff does not always seems a good use of the time at the moment, but it pays off big time in the future (proper documentation = More user & less issues to deal with on github)

alexandrebarachant · 2018-04-07T17:00:30Z

moabb/paradigms/motor_imagery.py

            # skip raw if no event found
            return

        # get interval
        if self.interval is None:
            tmin, tmax = dataset.interval
        else:
-            tmin, tmax = self.interval
+            tmin, tmax = [t + dataset.interval[0] for t in self.interval]


We should add a something to define only one part of the interval.
for example [1, None] will just start from 1 to the end of the MI period.
This usecase allow to cut out the cue that usually contain a lot of predictive power and bias the results.

What do you think of changing the input parameter interval to tmin and tmax. I can make the change

yes actually thats easier and you're right about the None. It would be great if you did that :)

okay, im doing it

alexandrebarachant · 2018-04-07T18:38:06Z

Okay, I made the change, but i'm not done yet. I will take this occasion to document all paradigms.
please, do not merge yet.

quick question. are SinglePass and FilterBank supposed to be used ? How do they differs from repectively MotorImagery and FilterBankMotorImagery ?

vinay-jayaram · 2018-04-07T18:43:08Z

I would say it only exists if people want paradigms for non-motor imagery (there's some word association/navigation/positive memories datasets out there) so they know if it's a single pass or filter bank paradigm.

vinay-jayaram · 2018-04-07T18:45:13Z

yes -- they don't implement used_events() so they are still abstract classes, and theres less of a chance of weird mistakes where someone inherits from FBMotorImagery or MotorImagery and forgets to update the datasets or something (we should incidentally remove the datasets property from BaseMotorImagery if it isn't gone already)

alexandrebarachant · 2018-04-07T19:29:02Z

Putting here a list of thing to do before merging. I will take care of them tomorrow

Review dataset documentation, make sure they render properly on the doc and have consistant citations format.
rename some datasets (BBCIEEG, Upperlimb, GigaDB) to have a citation-like name (e.g. Zouh2014)
split BBCIEEG dataset in two (dataset A and B), since motor imagery and mental arithmetic are recorded on separate session, it should not be possible to have them together.
properly document paradigm child class. make sure they render correctly. remove **kargs to make sure parameters shows up in the doc.
flake 8 should pass for datasets.

My goal is to be done with those dataset, and not comme back on them again.

vinay-jayaram · 2018-04-07T19:43:41Z

I added a function to tests/datasets that runs through a dataset and checks (kinda) that it conforms to the interface. We should make that better and document it so we have an easier time adding new datasets in the future.

I have a flake8 checker when I code but do you think we could increase the maximum line size? It's a bit of a pain at 80

alexandrebarachant · 2018-04-07T19:53:14Z

I get a lot of error E501 with your code. The number of char is actually limited to 79.

If you work alone, its fine to pick the length you want, but in our case we have to conform with what most people use.

alexandrebarachant · 2018-04-07T20:33:20Z

moabb/tests/datasets.py

@@ -4,6 +4,22 @@
 from moabb.datasets.fake import FakeDataset


+def _run_tests_on_dataset(d):


this is more or less a duplicate of what i have in tests/download.py.
this require to have a local copy of the datasets. I did not find a way to skip this from travis, so i did not add it to the test suit.

alexandrebarachant · 2018-04-08T22:00:42Z

Okay, i'm done, there is just one thing to do : document the paradigms.

I'm still unsure about our current list of paradigms.
The motorImagery is both multiclass and binary classif. Some algorithm, like CSP, are supposed to be 2 class only.
anyway, lets document this and merge.

vinay-jayaram · 2018-04-08T22:07:04Z

I'm not happy with the structure myself, but I don't think CSP as two-class is as much a problem, since it can always be used one vs rest. Plus it would be interesting to see how that actually performs, I think?

alexandrebarachant · 2018-04-08T22:11:38Z

This is actually not a problem because both MNE and pyriemann implementation of CSP handle multiclass. What i was saying is that we have something not ideal because pipeline are supposed to declare their list of compatible paradigm, and if we have a pipeline that is not compatible with multiclass, then we have an issue.
That is was why i originally implemented a TwoClassMotorImagery and a MultiClassMotorImagery paradigm.

vinay-jayaram · 2018-04-08T22:23:15Z

It wouldn't be hard to do that now either (just change LeftRight) but I don't like where it's heading, there are just so many possible ways of structuring an analysis...

could we force the pipelines to specify their requirements and not the exact paradigm name, and then have each paradigm generate a 'requirements met' list from its arguments? E.g. instead of calling FilterBank(...) just have flag 'multiple_bandpass' as True or something? I'd say that's another PR though

alexandrebarachant · 2018-04-08T22:29:13Z

I agree, there is just too many possible combination here. I think you are right and we need to start having a more structured requirement list on the pipeline side instead of declaring the paradigm.

This is another PR.
Okay. lets merge this one an i will revisit the paradigm structure another day.

Vinay Jayaram added 3 commits April 5, 2018 13:05

Commit 1: weibo 2014 and zhou 2016 added

555ba57

Many changes also detailed in pull request

Fix in Weibo downloading code

cba8318

alexandrebarachant reviewed Apr 5, 2018

View reviewed changes

Vinay Jayaram added 3 commits April 5, 2018 16:34

Working version, all changes not yet done

f4339bf

Fixes

1e2d8e3

1. Use new function used_events to choose events for a paradigm within the paradigm 2. Fixed bug in Physionet where hand and feet imagery couldn't be used together 3. Added documentation to new datasets and put them in .rst file 4. Simplified download code in Weibo2014

cosmetic fix

63722f5

fixed error in new weibo dl code

f08009a

alexandrebarachant reviewed Apr 5, 2018

View reviewed changes

alexandrebarachant reviewed Apr 6, 2018

View reviewed changes

alexandrebarachant added 2 commits April 5, 2018 22:32

replace BaseMotorImagery from tutorial

1a3aab6

better doc and minor fix

650416b

alexandrebarachant reviewed Apr 6, 2018

View reviewed changes

Vinay Jayaram added 4 commits April 6, 2018 12:56

Added raw documentation to all datasets and removed trial_interval

a9c1205

Needs to be prettied up at some point...

Updated dataset_search to include interval param

a0070bd

Updated utils to use interval, fixes

fa0222f

- use on_missing - intervals specified by paradigm are relative to imagery start - added min interval search to utils - paradigm datasets property now use interval search

hotfix, interval in utils is float

d03ab09

alexandrebarachant reviewed Apr 6, 2018

View reviewed changes

Vinay Jayaram added 2 commits April 6, 2018 16:19

Fixed tutorial? and halfhearted doc changes

ed0972c

fixed filterbank script

1be88be

also fixed bug in ImageryNClass init

fixed failing tests

9142d54

BaseMotorImagery is still abstract

alexandrebarachant reviewed Apr 7, 2018

View reviewed changes

alexandrebarachant approved these changes Apr 7, 2018

View reviewed changes

alexandrebarachant reviewed Apr 7, 2018

View reviewed changes

alexandrebarachant added 2 commits April 7, 2018 14:05

change interval to tmin, tmax

ed7e112

wip

741fbaa

alexandrebarachant reviewed Apr 7, 2018

View reviewed changes

alexandrebarachant added 3 commits April 8, 2018 17:19

wip

a4c9c36

small fix

737e744

fix tests

05347fe

vinay-jayaram merged commit 9ad04c3 into master Apr 8, 2018

vinay-jayaram deleted the new_datasets branch April 8, 2018 22:33

		FILES.append('https://dataverse.harvard.edu/api/access/datafile/2499179')


		def eeg_data_path(base_path, subject):



		class Weibo2014(BaseDataset):
		"""Weibo 2014 Motor Imagery dataset"""


		class Weibo2014(BaseDataset):
		"""Weibo 2014 Motor Imagery dataset [1]

		@@ -4,6 +4,22 @@
		from moabb.datasets.fake import FakeDataset


		def _run_tests_on_dataset(d):

New datasets and parameter confirmation #32

New datasets and parameter confirmation #32

Conversation

vinay-jayaram commented Apr 5, 2018

vinay-jayaram commented Apr 5, 2018

alexandrebarachant commented Apr 5, 2018

Choose a reason for hiding this comment

Choose a reason for hiding this comment

alexandrebarachant Apr 5, 2018 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

vinay-jayaram commented Apr 5, 2018 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

alexandrebarachant Apr 6, 2018 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

alexandrebarachant Apr 6, 2018 • edited Loading

Choose a reason for hiding this comment

vinay-jayaram Apr 6, 2018 • edited Loading

Choose a reason for hiding this comment

vinay-jayaram commented Apr 6, 2018

Choose a reason for hiding this comment

Choose a reason for hiding this comment

alexandrebarachant commented Apr 6, 2018 • edited Loading

vinay-jayaram commented Apr 6, 2018 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

alexandrebarachant left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

alexandrebarachant commented Apr 7, 2018

vinay-jayaram commented Apr 7, 2018

vinay-jayaram commented Apr 7, 2018

alexandrebarachant commented Apr 7, 2018 • edited Loading

vinay-jayaram commented Apr 7, 2018

alexandrebarachant commented Apr 7, 2018

Choose a reason for hiding this comment

alexandrebarachant commented Apr 8, 2018

vinay-jayaram commented Apr 8, 2018

alexandrebarachant commented Apr 8, 2018

vinay-jayaram commented Apr 8, 2018

alexandrebarachant commented Apr 8, 2018

alexandrebarachant Apr 5, 2018 •

edited

Loading

vinay-jayaram commented Apr 5, 2018 •

edited

Loading

alexandrebarachant Apr 6, 2018 •

edited

Loading

alexandrebarachant Apr 6, 2018 •

edited

Loading

vinay-jayaram Apr 6, 2018 •

edited

Loading

alexandrebarachant commented Apr 6, 2018 •

edited

Loading

vinay-jayaram commented Apr 6, 2018 •

edited

Loading

alexandrebarachant commented Apr 7, 2018 •

edited

Loading