PSLF: A PID Controller-incorporated Second-order Latent Factor Analysis Model for Recommender System

TL;DR

This paper introduces PSLF, a novel latent factor model for recommender systems that incorporates PID controllers and second-order information to improve convergence and performance on high-dimensional, incomplete data.

Contribution

The paper proposes a PID controller-incorporated second-order latent factor model that enhances convergence and accuracy in graph representation learning for complex interaction data.

Findings

Outperforms four state-of-the-art models in convergence speed.

Achieves better generalization on HDI datasets.

Utilizes Hessian-vector products for second-order insights.

Abstract

A second-order-based latent factor (SLF) analysis model demonstrates superior performance in graph representation learning, particularly for high-dimensional and incomplete (HDI) interaction data, by incorporating the curvature information of the loss landscape. However, its objective function is commonly bi-linear and non-convex, causing the SLF model to suffer from a low convergence rate. To address this issue, this paper proposes a PID controller-incorporated SLF (PSLF) model, leveraging two key strategies: a) refining learning error estimation by incorporating the PID controller principles, and b) acquiring second-order information insights through Hessian-vector products. Experimental results on multiple HDI datasets indicate that the proposed PSLF model outperforms four state-of-the-art latent factor models based on advanced optimizers regarding convergence rates and generalization performance.