Release Notes - 2020.01.22

Welcome to the 2020.01.22 product release.  This article provides an overview of the product innovations and improvements delivered on 22 January 2020, 2019.  For all future release plans check out our release calendar for 2020.  The release notes are structured as follows:

User Experience

OneAi Innovations

Security Improvements

 Bugs, Performance & Platform Improvements

  • Data Ingestion Bugs Fixed

  • General Bugs Fixed and Minor Improvements

  • OneAi Bugs Fixed

User Experience

Update on Beta Release of New User Experience (UX) & Storyboards

  • Huge thank you to customers providing input from their testing of the new user experience.  We have had lots of super positive feedback and suggestions.  We have been smashing bugs and adding more features.  Connect with the support team for a fresh version of the release information presentation. 

One Ai Innovations

  • Added the ability to download Correlations from an EDA report. (ref 3597)

  • Updated the terminology in the Results Summary Screen to clearly display Model Coefficients when they are available on the fitted model object as a part of Feature Importance.  (ref 599)

  • The Results Summary screen now shows all hyper-parameter settings. Hyper-parameters refer to estimator/algorithm tuning parameters, e.g. if you use weighted classes when using a random forest.  Previously we only displayed the algorithm configurations the user changed vs. showing the ones One AI changed or were defaulted by the package used ( sklearn, yandex, etc ).  (ref 602)

  • One AI can optionally lump outlier values in a categorical variable into an 'other' value. The EDA report now shows the mapping of values to others when anomaly lumping is enabled. For example, if there are only 3 employees in the office locations of Austin and Dallas then these two values may get bucketed into 'other'. On the EDA Report an indicator will show in the variable status section letting users know that the value was bucketed into 'other'.  (ref 610)

  • Added enhanced messaging around low value counts in the label that can cause errors when using k-folds. (ref 234)

  • New user-friendly error messages have been added to One AI. The new messages include a clear explanation of what caused the error, what variable caused the error, and a prescription for fixing the error when available.  (ref 639) Example new error message:

"A negative value occurs in column employee.salary. The selected RobustScaler does not accept negative values. Please check employee.salary for validity and/or select another scaler."

Security Improvements

  • Users who access the One Model application via Single Sign-On (and don't have direct login credentials) sometimes select the "Forgot Password" link on the One Model login page, but as they can only use Single Sign-on there is no password to reset.  After this release, the user will now get an email telling them to use Single Sign-On along with a hyperlink.  (ref 3666)

  • It is now possible to download exports from the Exports page while proxying as a different user.  (ref 3122)

Bugs, Performance & Platform Improvements

Data Ingestion Bugs Fixed

  • Fixed an issue where unusual and unexpected configurations in Workday could cause the Workday  connector to fail.  (ref 3728)

  • On occasion the One Model connector for Greenhouse might fail if Greenhouse added tables to their API. We have enhanced the One Model connector to ensure greater resilience when this happens to avoid these issues in the future. (ref 2992)

  • Fixed an issue where Workday APIs would not be queried during the Workday Maintenance window. We implemented a change in-case the window finishes before the scheduled time to maximize the time available for data extraction for customers.  (ref 3854)

General Bugs Fixed and Minor Improvements

  • Fixed a minor issue where the theme color for the Data button in the Sidebar might not be correctly applied.(ref 3836)

  • Fixed a bug where the session timeout was being too strictly applied in some circumstances and not recognizing user interactions, e.g. running queries in Explore. (ref 3673)

One Ai Bugs Fixed

  • Fixed a bug that prevented users from conducting classifications against a numeric label. (ref 596)

Was this article helpful?

0 out of 0 found this helpful

Comments

0 comments

Please sign in to leave a comment.