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
[MRG] added variation_stop parameter to multylayer perceptron #6518
Conversation
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.
Can you please fix the doctest failure?
Count of interations to attempt before stopping if score | ||
is not improving on the train set or on the validation fraction | ||
if early_stopping is True. | ||
Shold be 2 or more |
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.
*should
Count of interations to attempt before stopping if score | ||
is not improving on the train set or on the validation fraction | ||
if early_stopping is True. | ||
Shold be 2 or more |
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.
should
@@ -1152,6 +1163,12 @@ class MLPRegressor(BaseMultilayerPerceptron, RegressorMixin): | |||
early stopping. Must be between 0 and 1. | |||
Only used if early_stopping is True | |||
|
|||
variation_stop : int, optional, default 2 | |||
Count of interations to attempt before stopping if score |
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.
iterations
@amueller Done |
A test to ensure:
would be useful. |
I think #9457 makes this redundant. |
To the issue #6512
As described in book of Simon Haykin "Neural Networks - A Comprehensive Foundation", pp. 237-240, we should find an early stopping point if we use a validation set. But curve of a score on validation set may have it's local minimums and we will end falling in them if we make only 2 iteration after it before stopping. So i added count of iterations made before stop after best score as a parameter. Got an improvemen on MNIST in a few percent with 50 iterations.