The Problem with Metrics is a Fundamental Problem for AI

Optimizing a given metric is a central aspect of most current AI approaches, yet overemphasizing metrics leads to manipulation, gaming, a myopic focus on short-term goals, and other unexpected negative consequences. This poses a fundamental contradiction for AI development. Through a series of real-world case studies, we look at various aspects of where metrics go wrong in practice and aspects of how our online environment and current business practices are exacerbating these failures. Finally, we propose a framework towards mitigating the harms caused by overemphasis of metrics within AI by: (1) using a slate of metrics to get a fuller and more nuanced picture, (2) combining metrics with qualitative accounts, and (3) involving a range of stakeholders, including those who will be most impacted.

Classification of Single-lead Electrocardiograms: TDA Informed Machine Learning

Atrial Fibrillation is a heart condition characterized by erratic heart rhythms caused by chaotic propagation of electrical impulses in the atria, leading to numerous health complications. State-of-the-art models employ complex algorithms that extract expert-informed features to improve diagnosis. In this note, we demonstrate how topological features can be used to help accurately classify single lead electrocardiograms. Via delay embeddings, we map electrocardiograms onto high-dimensional point-clouds that convert periodic signals to algebraically computable topological signatures. We derive features from persistent signatures, input them to a simple machine learning algorithm, and benchmark its performance against winning entries in the 2017 Physionet Computing in Cardiology Challenge.

Multiclass Total Variation Clustering

Ideas from the image processing literature have recently motivated a new set of clustering algorithms that rely on the concept of total variation. While these algorithms perform well for bi-partitioning tasks, their recursive extensions yield unimpressive results for multiclass clustering tasks. This paper presents a general framework for multiclass total variation clustering that does not rely on recursion. The results greatly outperform previous total variation algorithms and compare well with state-of-the-art NMF approaches.