New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
add default option from param_range in validation curve #129
Conversation
autoreject/autoreject.py
Outdated
@@ -84,7 +98,11 @@ def validation_curve(epochs, y, param_name, param_range, cv=None): | |||
param_name="thresh", param_range=param_range, | |||
cv=cv, n_jobs=1, verbose=0) |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Since we're at it, perhaps let's expose the b_jobs param.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
what's that? :)
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
probably the n_jobs
parameter for the validation_curve
function: Currently hardcoded to 1, see line 85(old)/99(new)
@agramfort CIs are failing ... maybe because of your French keyboard |
This looks like it could go into 0.2 easily. @agramfort are you done here? If so, I can help rebase and merge it. |
yes. Can you take over though?
… |
3525584
to
bad58ed
Compare
Codecov Report
@@ Coverage Diff @@
## master #129 +/- ##
==========================================
+ Coverage 92.02% 92.67% +0.65%
==========================================
Files 10 10
Lines 1178 1187 +9
==========================================
+ Hits 1084 1100 +16
+ Misses 94 87 -7 |
Rebased, @agramfort anything else? |
The values of the parameter that will be evaluated. | ||
If None, 15 values between the min and the max threshold | ||
will be tested. |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
The problem doing this is that you will have np.inf
as the error when the candidate threshold is the min threshold. Even if you do something like min threshold + epsilon (so that at least one epoch is retained), you have the same problem as for some folds it could end up being the min threshold
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
problem with None param or with a my custom grid?
Your custom grid. It will give np.inf score when no epochs are kept.
On Sat 8 Jun 2019 at 13:12, Alexandre Gramfort ***@***.***> wrote:
***@***.**** commented on this pull request.
------------------------------
In autoreject/autoreject.py
<#129 (comment)>:
> The values of the parameter that will be evaluated.
+ If None, 15 values between the min and the max threshold
+ will be tested.
problem with None param or with a my custom grid?
—
You are receiving this because you commented.
Reply to this email directly, view it on GitHub
<#129?email_source=notifications&email_token=ADY6FISFK2OG4523KNHXPGDPZOHZ7A5CNFSM4GY7DOPKYY3PNVWWK3TUL52HS4DFWFIHK3DMKJSXC5LFON2FEZLWNFSXPKTDN5WW2ZLOORPWSZGOB27HVRQ#discussion_r291803474>,
or mute the thread
<https://github.com/notifications/unsubscribe-auth/ADY6FIX7T672ELDIQ4SZLB3PZOHZ7ANCNFSM4GY7DOPA>
.
--
Sent from my iPhone
|
can you make a test so I see if we catch this nicely?
… |
added a test for you that fails |
if you get a loglik of -inf then it means the model is bad and to me it's not too much of a problem as you want to find the argmax of test_scores (so -inf will be ignored). do I miss something? |
sure. But when you plot with np.inf then you have a ton of whitespace ... |
if the plotting is done with your own matplotlib code is it our problem?
matplotlib just not plots the points as
if they were NaN? I can live with that
… |
3eca313
to
c3a2291
Compare
this is all I needed to know ;-) |
No description provided.