Advances in Machine Learning Interpretability and Efficiency

The field of machine learning is moving towards developing more efficient and interpretable models. Recent research has focused on improving active learning methods, which enable models to selectively request labels for unlabeled data, reducing the need for large amounts of labeled data. Additionally, there has been a surge in developing methods for estimating the influence of individual training examples on model behavior, which is crucial for model debugging and data curation. Noteworthy papers in this area include the introduction of Partial Batch Label Sampling for efficient active learning and the development of f-INE, a hypothesis testing framework for estimating influence under training randomness. Furthermore, research on budget-constrained active learning and variable importance methods has shown promising results in improving model performance and robustness. Overall, these advances have the potential to significantly improve the efficiency and reliability of machine learning models.

Sources

Myopic Bayesian Decision Theory for Batch Active Learning with Partial Batch Label Sampling

f-INE: A Hypothesis Testing Framework for Estimating Influence under Training Randomness

Z0-Inf: Zeroth Order Approximation for Data Influence

Influence Dynamics and Stagewise Data Attribution

Budget-constrained Active Learning to Effectively De-censor Survival Data

Doctor Rashomon and the UNIVERSE of Madness: Variable Importance with Unobserved Confounding and the Rashomon Effect

Balancing Performance and Reject Inclusion: A Novel Confident Inlier Extrapolation Framework for Credit Scoring

Assessing the robustness of heterogeneous treatment effects in survival analysis under informative censoring

Built with on top of