Paper Info
Title
Distill-VQ: Learning Retrieval Oriented Vector Quantization By Distilling Knowledge from Dense Embeddings
Abstract
Vector quantization (VQ) based ANN indexes, such as Inverted File System (IVF) and Product Quantization (PQ), have been widely applied to embedding based document retrieval thanks to the competitive time and memory efficiency. Originally, VQ is learned to minimize the reconstruction loss, i.e., the distortions between the original dense embeddings and the reconstructed embeddings after quantization. Unfortunately, such an objective is inconsistent with the goal of selecting ground-truth documents for the input query, which may cause severe loss of retrieval quality. Recent works identify such a defect, and propose to minimize the retrieval loss through contrastive learning. However, these methods intensively rely on queries with ground-truth documents, whose performance is limited by the insufficiency of labeled data. In this paper, we propose Distill-VQ, which unifies the learning of IVF and PQ within a knowledge distillation framework. In Distill-VQ, the dense embeddings are leveraged as "teachers'', which predict the query's relevance to the sampled documents. The VQ modules are treated as the "students'', which are learned to reproduce the predicted relevance, such that the reconstructed embeddings may fully preserve the retrieval result of the dense embeddings. By doing so, Distill-VQ is able to derive substantial training signals from the massive unlabeled data, which significantly contributes to the retrieval quality. We perform comprehensive explorations for the optimal conduct of knowledge distillation, which may provide useful insights for the learning of VQ based ANN index. We also experimentally show that the labeled data is no longer a necessity for high-quality vector quantization, which indicates Distill-VQ's strong applicability in practice. The evaluations are performed on MS MARCO and Natural Questions benchmarks, where Distill-VQ notably outperforms the SOTA VQ methods in Recall and MRR. Our code is avaliable at https://github.com/staoxiao/LibVQ.
Year
DOI
Venue
2022
10.1145/3477495.3531799
SIGIR '22: Proceedings of the 45th International ACM SIGIR Conference on Research and Development in Information Retrieval
Keywords
DocType
Citations 
Vector Quantization, Knowledge Distillation, Embedding Based Retrieval, Approximate Nearest Neighbour Search
Conference
0
PageRank 
References 
Authors
0.34
7
13
Name
Order
Citations
PageRank
Shitao Xiao102.03
Zheng Liu28410.09
Weihao Han321.46
Jianjin Zhang461.88
Defu Lian575946.15
Yeyun Gong69416.67
Qi Chen700.34
Fan Yang800.34
Hao Sun902.03
Yingxia Shao1021324.25
Denvy Deng1101.01
Qi Zhang12931179.66
Xing Xie139105527.49