Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Ada-Retrieval: An Adaptive Multi-Round Retrieval Paradigm for Sequential Recommendations

Lei Li, Jianxun Lian, Xiao Zhou, Xing Xie

TL;DR

Ada-Retrieval is proposed, an adaptive multi-round retrieval paradigm for recommender systems that iteratively refines user representations to better capture potential candidates in the full item space and maintains a model-agnostic design.

Abstract

Retrieval models aim at selecting a small set of item candidates which match the preference of a given user. They play a vital role in large-scale recommender systems since subsequent models such as rankers highly depend on the quality of item candidates. However, most existing retrieval models employ a single-round inference paradigm, which may not adequately capture the dynamic nature of user preferences and stuck in one area in the item space. In this paper, we propose Ada-Retrieval, an adaptive multi-round retrieval paradigm for recommender systems that iteratively refines user representations to better capture potential candidates in the full item space. Ada-Retrieval comprises two key modules: the item representation adapter and the user representation adapter, designed to inject context information into items' and users' representations. The framework maintains a model-agnostic design, allowing seamless integration with various backbone models such as RNNs or Transformers. We perform experiments on three widely used public datasets, incorporating five powerful sequential recommenders as backbone models. Our results demonstrate that Ada-Retrieval significantly enhances the performance of various base models, with consistent improvements observed across different datasets. Our code and data are publicly available at: https://github.com/ll0ruc/Ada-Retrieval.

Ada-Retrieval: An Adaptive Multi-Round Retrieval Paradigm for Sequential Recommendations

TL;DR

Ada-Retrieval is proposed, an adaptive multi-round retrieval paradigm for recommender systems that iteratively refines user representations to better capture potential candidates in the full item space and maintains a model-agnostic design.

Abstract

Retrieval models aim at selecting a small set of item candidates which match the preference of a given user. They play a vital role in large-scale recommender systems since subsequent models such as rankers highly depend on the quality of item candidates. However, most existing retrieval models employ a single-round inference paradigm, which may not adequately capture the dynamic nature of user preferences and stuck in one area in the item space. In this paper, we propose Ada-Retrieval, an adaptive multi-round retrieval paradigm for recommender systems that iteratively refines user representations to better capture potential candidates in the full item space. Ada-Retrieval comprises two key modules: the item representation adapter and the user representation adapter, designed to inject context information into items' and users' representations. The framework maintains a model-agnostic design, allowing seamless integration with various backbone models such as RNNs or Transformers. We perform experiments on three widely used public datasets, incorporating five powerful sequential recommenders as backbone models. Our results demonstrate that Ada-Retrieval significantly enhances the performance of various base models, with consistent improvements observed across different datasets. Our code and data are publicly available at: https://github.com/ll0ruc/Ada-Retrieval.
Paper Structure (32 sections, 19 equations, 5 figures, 4 tables)

This paper contains 32 sections, 19 equations, 5 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Deep Retrieval
  4. Sequential Recommendation
  5. Preliminaries
  6. Problem Formulation
  7. Base Sequential Model
  8. Methodology
  9. Item Representation Adapter
  10. Learnable Filter Layer
  11. Context-aware Attention Layer
  12. A Case with Sequential User Model
  13. User Representation Adapter
  14. Gated Recurrent Unit Layer
  15. Multi-Layer Perceptron Layer
  16. ...and 17 more sections

Figures (5)

  • Figure 1: Illustrations of (a) the conventional single-round retrieval paradigm, and (b) our proposed adaptive multi-round retrieval paradigm, in which the final retrieval result is the union of each individual retrieval $K_i$.
  • Figure 2: An overview of the traditional retrieval model (a) and our proposed Ada-Retrieval paradigm (b), which consists of two key parts: the item representation adapter (c) and the user representation adapter (d). We use colored elements to indicate the new components in Ada-Retrieval.
  • Figure 3: Effect of the number of recommendation batches.
  • Figure 4: Sensitivity analysis of parameter $\lambda$.
  • Figure 5: Improvement between Ada-Retrieval and SASRec on each round.