Blog Posts
Connecting Interpretability and Robustness in Decision Trees through Separation
Trustworthy machine learning (ML) has emerged as a crucial topic for the success of ML models. This post focuses on three fundamental properties of trustworthy ML models -- high accuracy, interpretability, and robustness. Building on ideas from ensemble learning, we construct a tree-based model that is guaranteed to be adversely robust, interpretable, and accurate on linearly separable data. Experiments confirm that our algorithm yields classifiers that are both interpretable, robust, and have high accuracy.
Based on a paper published at ICML 2021
Explainable k-means Clustering
Popular algorithms for learning decision trees can be arbitrarily bad for clustering. We present a new algorithm for explainable clustering that has provable guarantees --- the Iterative Mistake Minimization (IMM) algorithm. This algorithm exhibits good results in practice. It's running time is comparable to k-means implemented in sklearn. So our method gives explanations basically for free. Our code is available on GitHub.
Based on: Explainable k-Means and k-Medians Clustering, ICML 2020
Explainable 2-means Clustering: Five Lines Proof
In this post we show why only one feature is enough to define a good 2-means clustering. And we do it using only 5 inequalities (!)
Based on: Explainable k-Means and k-Medians Clustering, ICML 2020