BIU Learning Club, January 27, 2025: Statistics-Powered ML: Building Trust and Robustness in Black-Box Predictions

Modern ML models produce valuable predictions across various applications, influencing people’s lives, opportunities, and scientific advancements. However, these systems can fail in unexpected ways, delivering unreliable inferences and perpetuating biases present in the data. This issue is particularly pronounced in high-stakes applications, where models are trained on increasingly diverse, incomplete, and noisy data and then deployed in dynamic environments—conditions that often exacerbate test-time failures.

This talk navigates two key questions: How can fundamental statistical principles be leveraged to rigorously build trust in modern ML systems? And how can these principles inspire the development of robust learning paradigms?

In the first part, I will discuss recent advances in conformal prediction—a statistical wrapper for any black-box model that provides precise error bounds on ML predictions. I will focus on scenarios where training data is corrupted or biased, such as through missing features and labels, and introduce a framework for constructing predictive uncertainty estimates that remain valid despite distribution shifts between the corrupted and unknown clean data.

In the second part, I will show how sequential statistical testing can enable a novel test-time training scheme, allowing a pre-trained model to adapt online to unfamiliar environments. For instance, consider an image classification task where test images are captured under varying illumination conditions that differ from the training setup. Building on conformal betting martingales, I will first introduce a monitoring tool to detect data drifts. Using this tool, I will derive a rigorous ‘anti-drift correction’ mechanism grounded in (online) optimal transport principles. This mechanism forms the foundation of a self-training scheme that produces robust predictions invariant to dynamically changing environments.