MRG: Add add_noise to simulation

[larsoner]

One part of what we need for #5058 (comment)

[larsoner]

@agramfort done

[codecov]

Codecov Report

Merging #5859 into master will increase coverage by 0.06%.
The diff coverage is 98.64%.

@@            Coverage Diff             @@
##           master    #5859      +/-   ##
==========================================
+ Coverage   88.62%   88.69%   +0.06%     
==========================================
  Files         373      396      +23     
  Lines       69355    71871    +2516     
  Branches    11669    12071     +402     
==========================================
+ Hits        61465    63744    +2279     
- Misses       5037     5199     +162     
- Partials     2853     2928      +75

[agramfort]

ok for you @makkostya and @ngayraud ?

[agramfort]

ping @makkostya @ngayraud @sdeslauriers

ok for you?

[ngayraud]

Looks good!

[agramfort]

great

ok for you too @makkostya ?

[makkostya]

If I understood well, the function is replacing the inst.data by simulated noise, not adding it?
Thus adding noise is done outside the function (something like soi.data + noise.data)?
If it is like this, I'm Ok with it.

Just the line:
for epoch in data:
will it work properly if inst is a raw?

[larsoner]

If I understood well, the function is replacing the inst.data by simulated noise, not adding it?

Oops, that's a bug.

will it work properly if inst is a raw?

Yes internally the function makes a new view of the data array, i.e. shape (1, 306, ...) if it's raw so it's unified.

[agramfort]

Oops, that's a bug.

good catch then :) test seems not good enough :-)

[larsoner]

Agreed :)

[ngayraud]

@makkostya good catch!

[agramfort]

thx @larsoner and @makkostya / @ngayraud for taking a look.

it would be great to have a tutorial soon on how to simulate data.