Hypothesis Class

Hypothesis classes, sets of possible models used in machine learning, are central to understanding learnability and generalization performance. Current research focuses on characterizing learnability for various hypothesis classes, including those used in multiclass classification, regression, and active learning, often analyzing sample complexity and algorithmic efficiency in both realizable and agnostic settings. This involves developing new theoretical tools like variations of VC dimension and exploring the impact of factors such as model complexity, data dependence, and the type of feedback received. These advancements refine our understanding of learning algorithms' capabilities and limitations, informing the design of more efficient and robust machine learning systems.

Papers

November 28, 2022

PAC Verification of Statistical Algorithms
Saachi Mutreja, Jonathan Shafer
Statistical Learning PAC Learning Hypothesis Class Agnostic PAC

November 16, 2022

Comparative Learning: A Sample Complexity Theory for Two Hypothesis Classes
Lunjia Hu, Charlotte Peale
Sample Complexity PAC Learning Hypothesis Class Comparative Learning

November 7, 2022

A Characterization of List Learnability
Moses Charikar, Chirag Pabbaraju
New Characterization PAC Learnability Hypothesis Class K$ Distance

August 31, 2022

Fine-Grained Distribution-Dependent Learning Curves
Olivier Bousquet, Steve Hanneke, Shay Moran, Jonathan Shafer, Ilya Tolstikhin
Fine Grained Learning Rate PAC Learning Hypothesis Class PAC Bayes Bound Learning Curve

April 26, 2022

An Empirical Study of the Occurrence of Heavy-Tails in Training a ReLU Gate
Sayar Karmakar, Anirbit Mukherjee
Empirical Study Deep Learning Algorithm Stochastic Way Rare Event Hypothesis Class Heavy Tail

November 24, 2021

Differentially Private Nonparametric Regression Under a Growth Condition
Noah Golowich
Private Algorithm Private Machine Learning Hypothesis Class Private Linear Regression Growth Monitoring