Search

Embedding retrieval aims to learn a shared semantic representation space for both queries and items, thus enabling efficient and effective item retrieval using approximate nearest neighbor (ANN) algorithms. In current industrial practice, retrieval systems typically retrieve a fixed number of items for different queries, which actually leads to insufficient retrieval (low recall) for head queries and irrelevant retrieval (low precision) for tail queries. Mostly due to the trend of frequentist approach to loss function designs, till now there is no satisfactory solution to holistically address this challenge in the industry. In this paper, we move away from the frequentist approach, and take a novel \textbf{p}robabilistic approach to \textbf{e}mbedding \textbf{b}ased \textbf{r}etrieval (namely \textbf{pEBR}) by learning the item distribution for different queries, which enables a dynamic cosine similarity threshold calculated by the probabilistic cumulative distribution function (CDF) value. The experimental results show that our approach improves both the retrieval precision and recall significantly. Ablation studies also illustrate how the probabilistic approach is able to capture the differences between head and tail queries.

Traditional self-interference cancellation (SIC) methods are common in full-duplex (FD) integrated sensing and communication (ISAC) systems. However, exploring new SIC schemes is important due to the limitations of traditional approaches. With the challenging limitations of traditional SIC approaches, this paper proposes a novel simultaneous transmitting and reflecting reconfigurable intelligent surface (STAR-RIS)-enabled FD ISAC system, where STAR-RIS enhances simultaneous communication and target sensing and reduces self-interference (SI) to a level comparable to traditional SIC approaches. The optimization of maximizing the sensing signal-to-interference-plus-noise ratio (SINR) and the communication sum rate, both crucial for improving sensing accuracy and overall communication performance, presents significant challenges due to the non-convex nature of these problems. Therefore, we develop alternating optimization algorithms to iteratively tackle these problems. Specifically, we devise the semi-definite relaxation (SDR)-based algorithm for transmit beamformer design. For the reflecting and refracting coefficients design, we adopt the successive convex approximation (SCA) method and implement the SDR-based algorithm to tackle the quartic and quadratic constraints. Simulation results validate the effectiveness of the proposed algorithms and show that the proposed deployment can achieve better performance than that of the benchmark using the traditional SIC approach without STAR-RIS deployment.

Reconfigurable intelligent surfaces (RISs) have emerged as a promising technology in wireless communications. Simultaneously transmitting and reflecting RIS (STAR-RISs) in particular have garnered significant attention due to their dual capabilities of simultaneous transmission and reflection, underscoring their potential applications in critical scenarios within the forthcoming sixth-generation (6G) technology landscape. Moreover, full-duplex (FD) systems have emerged as a breakthrough research direction in wireless transmission technology due to their high spectral efficiency. This paper explores the application potential of STAR-RIS in FD systems for future wireless communications, presenting an innovative technology that provides robust self-interference cancellation (SIC) capabilities for FD systems. We utilize the refraction functionality of STAR-RIS enhances the transmission capacity of FD systems, while its reflection functionality is used to eliminate self interference within the FD system. We delve into the applications of two different types of STAR-RIS in FD systems and compare their performance through simulations. Furthermore, we discuss the performance differences of STAR-RIS empowered FD systems under various configurations in a case study, and demonstrate the superiority of the proposed deep learning-based optimization algorithm. Finally, we discuss possible future research directions for STAR-RIS empowered FD systems., Comment: Accepted for publication in IEEE Communications Standards Magazine

Physical layer security (PLS) technologies are expected to play an important role in the next-generation wireless networks, by providing secure communication to protect critical and sensitive information from illegitimate devices. In this paper, we propose a novel secure communication scheme where the legitimate receiver use full-duplex (FD) technology to transmit jamming signals with the assistance of simultaneous transmitting and reflecting reconfigurable intelligent surface (STARRIS) which can operate under the energy splitting (ES) model and the mode switching (MS) model, to interfere with the undesired reception by the eavesdropper. We aim to maximize the secrecy capacity by jointly optimizing the FD beamforming vectors, amplitudes and phase shift coefficients for the ESRIS, and mode selection and phase shift coefficients for the MS-RIS. With above optimization, the proposed scheme can concentrate the jamming signals on the eavesdropper while simultaneously eliminating the self-interference (SI) in the desired receiver. To tackle the coupling effect of multiple variables, we propose an alternating optimization algorithm to solve the problem iteratively. Furthermore, we handle the non-convexity of the problem by the the successive convex approximation (SCA) scheme for the beamforming optimizations, amplitudes and phase shifts optimizations for the ES-RIS, as well as the phase shifts optimizations for the MS-RIS. In addition, we adopt a semi-definite relaxation (SDR) and Gaussian randomization process to overcome the difficulty introduced by the binary nature of mode optimization of the MS-RIS. Simulation results validate the performance of our proposed schemes as well as the efficacy of adapting both two types of STAR-RISs in enhancing secure communications when compared to the traditional selfinterference cancellation technology., Comment: 12 pages, 7 figures

Retrieval augmentation, which enhances downstream models by a knowledge retriever and an external corpus instead of by merely increasing the number of model parameters, has been successfully applied to many natural language processing (NLP) tasks such as text classification, question answering and so on. However, existing methods that separately or asynchronously train the retriever and downstream model mainly due to the non-differentiability between the two parts, usually lead to degraded performance compared to end-to-end joint training. In this paper, we propose Differentiable Retrieval Augmentation via Generative lANguage modeling(Dragan), to address this problem by a novel differentiable reformulation. We demonstrate the effectiveness of our proposed method on a challenging NLP task in e-commerce search, namely query intent classification. Both the experimental results and ablation study show that the proposed method significantly and reasonably improves the state-of-the-art baselines on both offline evaluation and online A/B test., Comment: 5 pages, 2 figures; accepted by CIKM2023

BERT-style models pre-trained on the general corpus (e.g., Wikipedia) and fine-tuned on specific task corpus, have recently emerged as breakthrough techniques in many NLP tasks: question answering, text classification, sequence labeling and so on. However, this technique may not always work, especially for two scenarios: a corpus that contains very different text from the general corpus Wikipedia, or a task that learns embedding spacial distribution for a specific purpose (e.g., approximate nearest neighbor search). In this paper, to tackle the above two scenarios that we have encountered in an industrial e-commerce search system, we propose customized and novel pre-training tasks for two critical modules: user intent detection and semantic embedding retrieval. The customized pre-trained models after fine-tuning, being less than 10% of BERT-base's size in order to be feasible for cost-efficient CPU serving, significantly improve the other baseline models: 1) no pre-training model and 2) fine-tuned model from the official pre-trained BERT using general corpus, on both offline datasets and online system. We have open sourced our datasets for the sake of reproducibility and future works., Comment: 5 pages, 3 figures; accepted by CIKM2022

Product quantization (PQ) coupled with a space rotation, is widely used in modern approximate nearest neighbor (ANN) search systems to significantly compress the disk storage for embeddings and speed up the inner product computation. Existing rotation learning methods, however, minimize quantization distortion for fixed embeddings, which are not applicable to an end-to-end training scenario where embeddings are updated constantly. In this paper, based on geometric intuitions from Lie group theory, in particular the special orthogonal group $SO(n)$, we propose a family of block Givens coordinate descent algorithms to learn rotation matrix that are provably convergent on any convex objectives. Compared to the state-of-the-art SVD method, the Givens algorithms are much more parallelizable, reducing runtime by orders of magnitude on modern GPUs, and converge more stably according to experimental studies. They further improve upon vanilla product quantization significantly in an end-to-end training scenario., Comment: published in ICLR 2022

Graph convolution networks (GCN), which recently becomes new state-of-the-art method for graph node classification, recommendation and other applications, has not been successfully applied to industrial-scale search engine yet. In this proposal, we introduce our approach, namely SearchGCN, for embedding-based candidate retrieval in one of the largest e-commerce search engine in the world. Empirical studies demonstrate that SearchGCN learns better embedding representations than existing methods, especially for long tail queries and items. Thus, SearchGCN has been deployed into JD.com's search production since July 2020., Comment: 2 pages, 1 figure; accepted by SIGIR2021 industry track

Embedding index that enables fast approximate nearest neighbor(ANN) search, serves as an indispensable component for state-of-the-art deep retrieval systems. Traditional approaches, often separating the two steps of embedding learning and index building, incur additional indexing time and decayed retrieval accuracy. In this paper, we propose a novel method called Poeem, which stands for product quantization based embedding index jointly trained with deep retrieval model, to unify the two separate steps within an end-to-end training, by utilizing a few techniques including the gradient straight-through estimator, warm start strategy, optimal space decomposition and Givens rotation. Extensive experimental results show that the proposed method not only improves retrieval accuracy significantly but also reduces the indexing time to almost none. We have open sourced our approach for the sake of comparison and reproducibility., Comment: 4 pages, 4 figures; accepted by SIGIR2021

Result relevance scoring is critical to e-commerce search user experience. Traditional information retrieval methods focus on keyword matching and hand-crafted or counting-based numeric features, with limited understanding of item semantic relevance. We describe a highly-scalable feed-forward neural model to provide relevance score for (query, item) pairs, using only user query and item title as features, and both user click feedback as well as limited human ratings as labels. Several general enhancements were applied to further optimize eval/test metrics, including Siamese pairwise architecture, random batch negative co-training, and point-wise fine-tuning. We found significant improvement over GBDT baseline as well as several off-the-shelf deep-learning baselines on an independently constructed ratings dataset. The GBDT model relies on 10 times more features. We also present metrics for select subset combinations of techniques mentioned above., Comment: 6 pages