Bregman divergence
PulseAugur coverage of Bregman divergence — every cluster mentioning Bregman divergence across labs, papers, and developer communities, ranked by signal.
2 day(s) with sentiment data
-
New mirror descent algorithms with logarithmic barriers detailed in arXiv paper
This paper introduces new convergence guarantees for mirror descent and proximal mirror descent algorithms when employing a logarithmic barrier. The authors propose a method that avoids the Bregman divergence blowing up…
-
New V-ECE method improves estimation of probabilistic classification calibration errors
Researchers have introduced V-ECE, a novel method for estimating general expected calibration errors in probabilistic classification. Unlike previous methods that were limited to specific divergence measures or required…
-
Bregman Divergences Unified for Neural Network Optimizer Improvement
Researchers have developed a unified framework using Bregman divergences to analyze the impact of different divergence choices on neural network optimizers, specifically Shampoo. This framework connects various popular …
-
New method improves parameter estimation for discrete models
Researchers have developed a new method for estimating parameters in unnormalized discrete models by combining empirical localization with deformed Bregman divergences. This approach significantly reduces computational …
-
New Riesz Regression Framework Unifies Debiased Machine Learning
Researchers have introduced generalized Riesz regression, a novel framework designed to unify and improve debiased machine learning techniques. This new method utilizes Riesz representer fitting under Bregman divergence…
-
New Newton's Method Achieves O(1/k^3) Convergence Rate
Researchers have developed a novel direct accelerated Newton method for minimizing convex functions with Lipschitz continuous Hessians. This new algorithm operates solely with primal variables and requires only one line…
-
New Bayesian Framework Enhances Feature Extraction for Spatio-Temporal Data
Researchers have developed a new Bayesian feature extraction framework designed for high-dimensional spatio-temporal data, particularly useful in scientific domains. This framework utilizes Gaussian and Diffused-gamma p…
-
New framework expands calibeating to general proper losses
Researchers have developed a new framework for calibeating using regret minimization, extending previous work on specific loss functions to a broader family of proper losses. This approach utilizes Bregman divergences t…