empirical risk minimization
Empirical risk minimization (ERM) is a principle in statistical learning that aims to minimize the average loss on a training dataset. It serves as the foundation for many machine learning algorithms, guiding them towards generalizable performance across unseen data.
- Bivariate Matrix-valued Linear Regression (BMLR): Finite-sample performance under Identifiability and Sparsity Assumptions
- Discretization-free Multicalibration through Loss Minimization over Tree Ensembles
- On Agnostic PAC Learning in the Small Error Regime
- Prediction-Powered Causal Inferences
- Probably Approximately Precision and Recall Learning
- Semi-Supervised Regression with Heteroscedastic Pseudo-Labels
- The Nuclear Route: Sharp Asymptotics of ERM in Overparameterized Quadratic Networks
- Tight Generalization Bounds for Large-Margin Halfspaces
- Tradeoffs between Mistakes and ERM Oracle Calls in Online and Transductive Online Learning
- Which Algorithms Have Tight Generalization Bounds?