Release Notes V1.21.0¶

Version	1.21.0
Release Date	30^th April 2023

Data and Model Management

Experiment Revamp:
- Added a new module called Algorithm which allows users to register algorithms for model training. The module provides a simple interface for users to register their own standard and approved algorithms, making it easier for them to integrate their own custom code into the platform.
- Enhanced the existing Dataset Module, which now allows users to register algorithm-specific datasets. This new capability provides greater flexibility and control over the data used for training models. Furthermore, the new Dataset Module now includes the ability to add Transforms, allowing users to preprocess their data before training and also to specify data inputs based on an Algorithm.
- Enhanced the Experiment Module, providing users with a more intuitive interface for defining experiments. The new interface enables users to easily choose which algorithm to run with which dataset. Additionally, users can now provide additional user inputs required in the Algorithm, giving them greater control over the model training process.
Eliminated the overhead of always creating a primary table for Table Key. A primary table would still be required for a Table Key which is mapped to a Platform Entity.
Added flexibility in providing hierarchy levels for Platform Entities where:
- Hierarchy levels can have gaps
- Multiple Platform Entities can have same hierarchy level
Model Transform simulations are made optional for a Model Approval.

Strategy Automation

Removed Model Suites in Policies: The creation of a model suite was previously required to use a model in a policy, but this additional step has now been eliminated.
Reason Codes for policy decisions: Ability to add Reason Codes for each rules and optimization strategy in a policy

Common

Reporting and Dashboards:
- Improved user experience when creating Dashboard Views**
- Implemented governance and versioning for Dashboard Views
- Moved Dashboard View registry from Settings to Resources
- Simplifying the structures of the data used for report creation
Job Execution
- Introduced Job Board which shows the list of jobs currently running and the jobs queued for execution. This will help users see their job’s position in the queue.
- Enhanced Job Runtime Popup to include information about Compiling, Queuing and Execution time. The total job runtime is also broken down to show the corresponding info.
- Reuse of already computed filtered data: Added the capability for the users to use filtered data tables from already run jobs when running a new job. This would remove the filtering step completely from a job, reducing the overall job run time by about 70% in most cases, especially when the input data tables are huge.
Notify users when a new version of an approved object that they are using downstream has been created. A notification will be sent out to all the users the downstream users of the object.
Add User related information in the corridor python package.

Technical