Brief

Researchers have developed a new, mathematically sound, and computationally efficient method for measuring model complexity. This approach, based on analyzing similarities in model gradients across different inputs, is applicable to a wide range of models, including parametric, non-parametric, and kernel-based types. The proposed measure unifies and generalizes existing complexity metrics for various models like decision trees and neural networks, offering new insights into phenomena such as double descent. AI

Researchers have developed a new local learning method for selecting covariates in causal effect estimation, bypassing the need for pretreatment or causal sufficiency assumptions. This approach identifies a local boundary to efficiently search for valid adjustment sets, improving computational efficiency. Experiments on synthetic and real-world data demonstrate accurate causal effect estimation with significant speedups. AI