PSD Representations for Effective Probability Models

Authors: Alessandro Rudi, Carlo Ciliberto

NeurIPS 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Theoretical In this work, we show that a recently proposed class of positive semidefinite (PSD) models for non-negative functions is particularly suited to this end. In particular, we characterize both approximation and generalization capabilities of PSD models, showing that they enjoy strong theoretical guarantees. Moreover, we show that we can perform efficiently both sum and product rule in closed form via matrix operations, enjoying the same versatility of mixture models. Our results open the way to applications of PSD models to density estimation, decision theory and inference.
Researcher Affiliation Academia Alessandro Rudi Inria, École normale supérieure, CNRS, PSL Research University, Paris, France alessandro.rudi@inria.fr Carlo Ciliberto Department of Computer Science University College London, London, UK c.ciliberto@ucl.ac.uk
Pseudocode Yes Algorithm 1 PSD Hidden Markov Model
Open Source Code No The paper states, 'we plan to develop a library for operations with PSD models and make it available to the community' in the 'Future Directions' section, indicating future availability, not current release, and no specific link is provided for the current work.
Open Datasets No The paper is theoretical and does not conduct experiments on specific, named datasets for which public availability information would be relevant.
Dataset Splits No The paper is theoretical and does not describe empirical experiments with training, validation, and test splits.
Hardware Specification No The paper is theoretical and does not report on empirical experiments, therefore no hardware specifications are mentioned.
Software Dependencies No The paper is theoretical and does not report on empirical experiments, therefore no specific software dependencies with version numbers are mentioned.
Experiment Setup No The paper is theoretical and does not describe an empirical experimental setup with hyperparameter values or training configurations.