publications

2025

1. TRIP: A Nonparametric Test to Diagnose Biased Feature Importance Scores

   Aaron Foote and Danny Krizanc

   Aug 2025

   Abs DOI

   Along with accurate prediction, understanding the contribution of each feature to the making of the prediction, i.e., the importance of the feature, is a desirable and arguably necessary component of a machine learning model. For a complex model such as a random forest, such importances are not innate – as they are, e.g., with linear regression. Efficient methods have been created to provide such capabilities, with one of the most popular among them being permutation feature importance due to its efficiency, model-agnostic nature, and perceived intuitiveness. However, permutation feature importance has been shown to be misleading in the presence of dependent features as a result of the creation of unrealistic observations when permuting the dependent features. In this work, we develop TRIP (\textbfTest for \textbfReliable \textbfInterpretation via \textbfPermutation), a test requiring minimal assumptions that is able to detect unreliable permutation feature importance scores that are the result of model extrapolation. To build on this, we demonstrate how the test can be complemented in order to allow its use in high dimensional settings. Through testing on simulated data and applications, our results show that the test can be used to reliably detect when permutation feature importance scores are unreliable.

2. Nonogram: Complexity of Inference and Phase Transition Behavior

   Aaron Foote and Danny Krizanc

   Aug 2025

   Abs DOI

   Nonogram is a popular combinatorial puzzle (similar in nature to Sudoku or Minesweeper) in which a puzzle solver must determine if there exists a setting of the puzzle parameters that satisfy a given set of constraints. It has long been known that the problem of deciding if a solution exists is a computationally difficult problem. Despite this fact, humans still seem to enjoy playing it. This work aims to reconcile these seemingly contradictory facts by (1) analyzing the complexity of the inference problem for Nonogram (the problem of determining if there exists a puzzle parameter that can be inferred from the constraints without guessing) and (2) experimentally establishing the existence of a phase transition behavior for this inference problem. Our results show that the difficulty of the inference problem is largely determined by the density of filled cells (positive parameters) in a given puzzle. Along the way we implement an efficient encoding of a Nonogram board as a Boolean formula in Conjunctive Normal Form (CNF) through the use of regular expressions in order to make our experiments feasible.

2023

1. Factors in Learning Dynamics Influencing Relative Strengths of Strategies in Poker Simulation

   Aaron Foote, Maryam Gooyabadi, and Nikhil Addleman

   Games, Aug 2023

   Abs DOI

   Poker is a game of skill, much like chess or go, but distinct as an incomplete information game. Substantial work has been done to understand human play in poker, as well as the optimal strategies in poker. Evolutionary game theory provides another avenue to study poker by considering overarching strategies, namely rational and random play. In this work, a population of poker playing agents is instantiated to play the preflop portion of Texas Hold’em poker, with learning and strategy revision occurring over the course of the simulation. This paper aims to investigate the influence of learning dynamics on dominant strategies in poker, an area that has yet to be investigated. Our findings show that rational play emerges as the dominant strategy when loss aversion is included in the learning model, not when winning and magnitude of win are of the only considerations. The implications of our findings extend to the modeling of sub-optimal human poker play and the development of optimal poker agents.