[Mne_analysis] GSOC 2017 Idea Improve the Decoding Module.

[akshay gupta]

Hello mentors,
Let me share my views that I have after following some issues, tutorials,
doc's etc. This project will involve clubing of different modules like
enhance decoding objects to follow sklearn API, performance improvement in
GAT.

Follow sklearn API:
Make GAT and EMS to use GridSearch and CrossValidation methods of Scikit
Learn. In order to optimize memory and CPU time scoring and storing of
y_pred_ do not follow scikit learn API which disallows GAT and EMS to use
those methods.
My idea is to design a wrapper aroung GridSearch and CrossValidation
methods. So, instead of calling these methods directly from sklearn, they
will be called by methods in MNE itself. Function of these wrappers
function will be to handle input parameters and output return value of GAT
and EMS to make them compatible with GridSearch and Cross Validation.

Next coming to GAT optmization:
Reference issue #2188
Idea is to initialize each classifier with coeff. found in preceding time
thus reducing CPU time. warm_start feature of classifier in scikit learn
can allow us to do so. It has some difficulty like-

-> Classifier is cloned before fit which prevents reusing the same
estimator.
-> Not all classifier in sklearn has warm_start feature.

It will be good if we divide this task into sub tasks and tackle each
hurdle one by one.


If everything goes right I will soon start writing pre-proposal. My request
is to correct me if I'm wrong in anything.

Suggestion are most welcomed. This is from my own understanding it will be
excellent if someone can point out work that are renaming to be done here.

Thank you
Akshay Gupta

On Sat, Mar 4, 2017 at 4:47 PM, akshay gupta <akshay0742 at gmail.com> wrote:

> Thanks for pointing out the issues. I will soon make my hand dirty with
> them.
>
> Regards
> Akshay Gupta
>
> On Mar 4, 2017 12:06 AM, "Alexandre Gramfort" <alexandre.gramfort at telecom-
> paristech.fr> wrote:
>
>> hi,
>>
>> thanks for you interest.
>>
>> I'd recommend to watch the project on github, read our tutorials (you'll
>> certainly
>> find typos or more).
>>
>> Maybe this :
>>
>> https://github.com/mne-tools/mne-python/issues/2828
>>
>> is a good start.
>>
>> This one:
>>
>> https://github.com/mne-tools/mne-python/issues/3201
>>
>> is also probably doable of you followed the tutorials.
>>
>> Best,
>> Alex
>>
>> On Fri, Mar 3, 2017 at 1:07 PM, Akshay Gupta <akshay0724 at gmail.com>
>> wrote:
>>
>>> Hello everyone,
>>>
>>> As I introduced my self earlier, I'm a pre-final year student majoring
>>> computer science from AMU, India. I have strong coding skills and
>>> proficient in Python, Cython and C++. I'm also a small contributor to
>>> Scikit Learn and know some of it's internal working.
>>> I am looking forward to take part in GSOC 2017, and I'm interested to
>>> work on idea- improve the decoding module.
>>> I have been looking into the code base and found that it is well
>>> structured that help me a lot in becoming comfortable with the code base.
>>> I will start contributing to MNE soon, but it will be excellent if
>>> anyone can point out some issue just for start. Some pointers will be
>>> really helpful to start working on this module.
>>> Lastly I apologize if I have been rude in any manner.
>>>
>>> Thanks
>>> Akshay Gupta
>>>
>>> _______________________________________________
>>> Mne_analysis mailing list
>>> Mne_analysis at nmr.mgh.harvard.edu
>>> https://mail.nmr.mgh.harvard.edu/mailman/listinfo/mne_analysis
>>>
>>>
>>> The information in this e-mail is intended only for the person to whom
>>> it is
>>> addressed. If you believe this e-mail was sent to you in error and the
>>> e-mail
>>> contains patient information, please contact the Partners Compliance
>>> HelpLine at
>>> http://www.partners.org/complianceline . If the e-mail was sent to you
>>> in error
>>> but does not contain patient information, please contact the sender and
>>> properly
>>> dispose of the e-mail.
>>>
>>>
>>
>> _______________________________________________
>> Mne_analysis mailing list
>> Mne_analysis at nmr.mgh.harvard.edu
>> https://mail.nmr.mgh.harvard.edu/mailman/listinfo/mne_analysis
>>
>>
>> The information in this e-mail is intended only for the person to whom it
>> is
>> addressed. If you believe this e-mail was sent to you in error and the
>> e-mail
>> contains patient information, please contact the Partners Compliance
>> HelpLine at
>> http://www.partners.org/complianceline . If the e-mail was sent to you
>> in error
>> but does not contain patient information, please contact the sender and
>> properly
>> dispose of the e-mail.
>>
>>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mail.nmr.mgh.harvard.edu/pipermail/mne_analysis/attachments/20170313/e692168d/attachment.html