[FIX] Model's data_to_model_domain supports sparse matrix #4081
Conversation
PrimozGodec
force-pushed
the
fix-predictor
branch
2 times, most recently
from
d1ff14c
to
3e157b2
Compare
Codecov Report
@@ Coverage Diff @@
## master #4081 +/- ##
==========================================
+ Coverage 85.55% 85.56% +<.01%
==========================================
Files 385 385
Lines 69162 69178 +16
==========================================
+ Hits 59175 59191 +16
Misses 9987 9987 |
Orange/base.py
Outdated
| @@ -320,9 +321,12 @@ def data_to_model_domain(): | |||
|
|
|||
| if self.original_domain.attributes != data.domain.attributes \ | |||
| and data.X.size \ | |||
| and not np.isnan(data.X).all(): | |||
| and not np.isnan( | |||
There was a problem hiding this comment.
I don't think this is right. You want to check if all the entries are not nan, but then you only check the non-zero entries for sparse matrices. This behaviour is not consistent with what will happen for dense matrices. Also, consider taking a look at potentially useful sparse utilities here and use those where possible. If a use case is missing, such as the one you're introducing right now, please add it there, otherwise, there's a lot of duplication.
There was a problem hiding this comment.
You are right. :) I totally forgot that in sparse matrix unknown values are actually zeroes. Now it is fixed. I implemented that as a function similar to any_nan.
PrimozGodec
force-pushed
the
fix-predictor
branch
3 times, most recently
from
9b0b67b
to
44f585a
Compare
PrimozGodec
force-pushed
the
fix-predictor
branch
from
44f585a
to
7d57895
Compare
Issue
Fixes #3987
Description of changes
Model's data_to_model_domain fixed to accept also sparse data.
Includes