DPPLikelihood

Property	Data
Created	2023-02-21
Updated	2023-03-03
Author	@Aiden
Tags	#study

Determinantal Point Process Likelihoods for Sequential Recommendation

Title	Venue	Year	Code
Determinantal Point Process Likelihoods for Sequential Recommendation	SIGIR	'22	✓

Abstract

Component	Definition	Example
Problem Definition	Existing sequential recommendation techniques lack tailored loss functions that fit naturally into the practical application scenario of sequential recommender systems.	Such as cross-entropy and Bayesian Personalized Ranking (BPR) are widely used in the sequential recommendation area, but suffer from 2 inherent drawbacks: 1. the dependencies among elements of a sequence are overlooked in these loss formulations; 2. Instead of balancing accuracy (quality) and diversity, only generating accurate results has been over emphasized.
Proposed Solution	Two new loss functions based on the Determinantal Point Process (DPP) likelihood, which captures natural dependencies among temporal actions, and a quality vs. diversity decomposition of the DPP kernel to push beyond accuracy-oriented loss functions.	DPP Set Likelihood-Based Loss, Conditional DPP Set Likelihood-Based Loss
Experiment Result	The proposed loss functions show marked improvements over state-of-the-art sequential recommendation methods in both quality and diversity metrics in experiments on three real-world datasets.	-