Claudio Bruderer (Product Manager at Modulos)
The latest version of Modulos AutoML is out: v.0.3.3. With this version, we’re shipping some major new features to help you rapidly build machine learning models and deploy them to production.
Data Science Updates
Jupyter Notebook for Solutions
The first thing you want to do when you have a new trained machine learning model is to experiment with it. That’s why we’re including a Jupyter notebook with everything you need to try out your model straight away. The Jupyter notebook contains information on how to generate predictions with the Modulos Solution. It also serves as an introduction tool for the online- and the batch-clients.
Report for Solutions
Every Solution now includes a Report. This HTML document contains insights about the performance of the model in the Solution. The metrics in the Report depend on the details of the workflow and are designed to help you evaluate the Solution.
With datetime support, AutoML can now interpret a wide range of date and time formats in your data set. There are no additional steps necessary as AutoML automatically detects dates and timestamps. Those date and time inputs are used in the background of AutoML to extract more information out of the data. This creates even better Solutions for you, which are also more directly applicable to the problem at hand.
Automatic Workflow Pausing
When should you stop looking for a better machine learning Solution? In theory, never, as the search space is infinite. Infinite workflows are not particularly practical, so we are adding automated pausing as a feature. When your workflow does not produce a better score for 200 candidate Solutions, AutoML will automatically pause your workflow. You can of course always resume it if you want AutoML to continue the search.
User Interface Improvements
New User creation and Password Reset
We updated the user interface to make it easy and secure for users and admins to create new users and change their passwords.
We now visualize the feature extraction step on the details page on each workflow overview page. Previously, this step simply happened in the background, but is now shown to you for added transparency.
Other Improvements and Bug Fixes
We additionally made a range of other fixes and improvements to AutoML, which include:
- Users using the Safari browser experienced issues with the User Management sidebar. This has been fixed.
- The sidebar menu in the tutorials is now always extended by default. This makes it easier to navigate the tutorials.
- The Bayesian Optimizer is now the default optimizer option. You can however always choose a different optimizer, if your problem requires it.