Brief

Researchers have developed a new method for derandomizing PAC-Bayes generalization bounds, specifically for smooth loss functions. This approach aims to create high-probability bounds for deterministic predictors by leveraging the smoothness properties of both the loss function and the predictor class. The study details the cost of transitioning from a Gibbs predictor to a deterministic predictor at the posterior mean, linking it to the generalization gap of the Jensen gap class, and proposes a practical regularizer inspired by the theoretical framework. AI

Researchers have developed a new method for training neural networks that is more robust to errors in labeled data. This approach, called symmetrization of loss functions, theoretically guarantees better performance when dealing with noisy labels. The study introduces specific multi-class loss functions, including SGCE and alpha-MAE, which interpolate between existing methods and offer control over smoothness, showing competitive results on benchmarks. AI