feat: efficient evaluation loops #307
Merged
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Summary
This pull request introduces the EfficientEvaluationLoops class within the moa.evaluation package. This class provides functionality for executing experiments programmatically, bypassing the need for a GUI or CLI. It offers the same capabilities as running Tasks, such as EvaluatePrequential, but is intended for use directly from code. This allows users to seamlessly integrate this class with other Java codebases.
Key Features:
Programmatic Experiment Execution: Enables execution of experiments without requiring a graphical user interface (GUI) or command line interface (CLI).
Task-Based Evaluation: Provides the same functionality as Task-based evaluations (e.g., EvaluatePrequential) for ease of use in code.
MOA Learners Integration: Facilitates exploration and execution of MOA learners programmatically.
PrequentialResult Class: An inner class that encapsulates the results of various evaluation types, such as Prequential, Cumulative, or Windowed evaluations. It stores ground truth and prediction values for plotting or debugging purposes.
Implementation Details:
The EfficientEvaluationLoops class contains methods to calculate test-then-train metrics and prequential windowed metrics simultaneously.
The PrequentialResult inner class is designed to store the results of evaluations, with support for ground truth and prediction values.
Includes several example methods demonstrating different evaluation scenarios, which can serve as the basis for future tests.
Important Considerations:
This class was initially designed to offer efficient evaluation loops for MOA learners within the CapyMOA framework. Any modifications to this class should be made with caution, considering its integration with CapyMOA.
For more details, refer to the CapyMOA website: www.capymoa.org.
Example Usage:
The main method in the EfficientEvaluationLoops class provides various example scenarios demonstrating the usage of the class, including prequential evaluation edge cases, efficiency evaluations, and different types of evaluation methods (e.g., regression, windowed evaluation).
Testing and examples
Example scenarios included in the main method demonstrate the functionality and performance of the class. These examples can be adapted into formal tests to ensure the robustness and reliability of the implementation.