Skip to content

[SPARK-2552][MLLIB] stabilize logistic function in pyspark #1493

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.

Sign up for GitHub

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

Closed
mengxr wants to merge 1 commit into from
Closed

[SPARK-2552][MLLIB] stabilize logistic function in pyspark #1493

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
5 changes: 4 additions & 1 deletion python/pyspark/mllib/classification.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -63,7 +63,10 @@ class LogisticRegressionModel(LinearModel):
def predict(self, x):
_linear_predictor_typecheck(x, self._coeff)
margin = _dot(x, self._coeff) + self._intercept
prob = 1/(1 + exp(-margin))
if margin > 0:
prob = 1 / (1 + exp(-margin))
else:
prob = 1 - 1 / (1 + exp(margin))
Copy link
Contributor

@naftaliharris naftaliharris Jul 29, 2014

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Better would be

prob = exp(margin)/(1 + exp(margin))

because for very small probabilities this computation is essentially doing 1 - 1. For example:

>>> from math import exp
>>> margin = -40
>>> 1 - 1 / (1 + exp(margin))
0.0
>>> exp(margin)/(1 + exp(margin))
4.248354255291589e-18
>>>

Copy link
Contributor Author

@mengxr mengxr Jul 30, 2014

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yes, that is definitely better. Could you submit a PR? We don't need a JIRA for small changes. Btw, please cache exp(margin) instead of computing it twice.

Copy link
Contributor

@naftaliharris naftaliharris Jul 30, 2014

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Sure, pull request here! #1652

Thanks a lot! :-)

return 1 if prob > 0.5 else 0


Expand Down