Posts

Predictive Equivalence in Decision Trees

Image
This blog post is based on joint work with Hayden McTavish and Jon Donnelly and is presented in our ICML 2025 paper , "Leveraging Predictive Equivalence in Decision Trees", advised by Dr.'s Margo Seltzer and Cynthia Rudin. A Top Down View of Decision Trees Figure 1: Two decision trees which encode precisely the same logical expression. Splits to the right correspond to the variable being True, left to False. Decision trees are generally accepted as a very interpretable model class for understanding tabular machine learning problems. Look through a classical machine learning textbook, and you'll find that algorithms for 'optimizing' decision trees are a dime a dozen -- Scikit-Learn implements CART, C4.5, C5.0, and ID3. Modern research (see GOSDT and SPLIT, STreeD, and DL8.5, among others) has taken these algorithms far beyond the old days of greedy optimization into the world of true optimality, subject to depth constraints and limits on the number of leaves in...
Post a Comment
Read more

What is Machine Unlearning?

Hello, world. This is my first blog post (ever)! In it, I will be exploring at an arm's length the concept of machine unlearning, and some broad areas of research in the field. I plan on following up on this summary article with technical reviews and dives into papers relating to the field, following my own trajectory of learning. I will also write about my own ideas about the field as they come to me! Here goes nothing... Machine Unlearning Why should I care?    In order to be compliant with recent trends in data privacy legislation, such as GDPR (General Data Protection Regulation) and the Virginia CDPA (Consumer Data Privacy Act), companies which store user data must delete said data by request of the user. This presents two immediate problems to such a company: first, if data is de-identified or anonymized, it may be difficult or impossible to track down all of a user's data in the company's databases; second, data is baked into machine learning models underpinning much...
1 comment
Read more