Roadmap

Anyone is welcomed to work on any feature or enhancement listed below. Your contributions would be clearly posted in Contributors and Changelog.

Long-term Objective

• Major Feature API Change Verify the effectiveness of new ensembles that could be included in torchensemble

• Efficiency Improve the training and evaluating efficiency of existing ensembles

New Functionality

• Major Feature API Change Add the partial-fit mode for all ensembles

• Major Feature API Change Support manually-specified evaluation metrics when calling predict()

• Major Feature API Change Integration with torch.utils.tensorboard for better visualization

• Major Feature API Change Support arbitrary training criteria for existing ensembles

Maintenance Related

• Enhancement Refactor existing code to reduce the redundancy

• Enhancement Reduce the randomness in the experiment part by running each experiment multiple times with different random seeds