-
Notifications
You must be signed in to change notification settings - Fork 270
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
Version Update to PyPi #8
Conversation
The codebase is completely cleaned up and separated into sub-modules in the way that makes most sense from the future development standpoint.
- added several datasets (cervical cancer, titanic, iris) - added a multi-classification model - added to spear_reductiong handling and user message for the case where previous interval only produce nan values (usually means there is a nan in x) - removed some debuggin codes from scan
- added several new datasets - added BinaryPerformance for unified scoring - added the 'missing' Keras metrics fmeasure, etc. - added the ability to change the monitor for early stop - reporting focus now on summarizing the best and the worst - round results are now produced in a far more cleaner way that allow using any metric, and also cases where acc and loss are not used - added saving of model (not yet saving the best model though) - cleaned up the template from redundant items - reduced dependy to template.py in a push to make it entire redudant
- renamed same files - now performance metric is unified for binary, multi-label, and multi-class - spear allows any reduction_metric - added pred_class() which automatically detects the type of prediction challenge - cleaned up results
- Improved the way that reporting handles change in labels - Made templates.py completely redundant as part of the wider effort to make the entire pipeline completely dynamic in the sense of not relying at all in string labels - Fixed a bug in spear_reducer that locked the changes to learning rate only - Changing the name to Talos
- added shapes - exposed all Dense layer parameters from hidden_layers() layer generator - changed data folder name to examples - cleaned up the namespace - fixed issues in the Reporting with the way the function names are handled - added a breast_cancer model to examples - added iris and breast_cancer params to examples - renamed the /examples to /notebooks
- there was an error being raised when no shape was set - the detector function was missing an import - results was in every case reporting the last epoch as peak result - early_stopper did not technically allow custom inputs for min_delta and patience
This will be part of the "generality" metric, so that talos optimizers will have to optimize against performance and generality at the same time, to avoid finding just high performance scores that will not generalize well.
- astetik plots are now available through ta.plots...
…ptimization on Keras with Breast Cancer Data.ipynb
Hello @mikkokotila! Thanks for submitting the PR.
|
No description provided.