Ziwei Fan

Ziwei Fan, Zhiwei Liu, Hao Peng et al. · salesforce

Self-supervised sequential recommendation significantly improves recommendation performance by maximizing mutual information with well-designed data augmentations. However, the mutual information estimation is based on the calculation of Kullback Leibler divergence with several limitations, including asymmetrical estimation, the exponential need of the sample size, and training instability. Also, existing data augmentations are mostly stochastic and can potentially break sequential correlations with random modifications. These two issues motivate us to investigate an alternative robust mutual information measurement capable of modeling uncertainty and alleviating KL divergence limitations. To this end, we propose a novel self-supervised learning framework based on Mutual WasserStein discrepancy minimization MStein for the sequential recommendation. We propose the Wasserstein Discrepancy Measurement to measure the mutual information between augmented sequences. Wasserstein Discrepancy Measurement builds upon the 2-Wasserstein distance, which is more robust, more efficient in small batch sizes, and able to model the uncertainty of stochastic augmentation processes. We also propose a novel contrastive learning loss based on Wasserstein Discrepancy Measurement. Extensive experiments on four benchmark datasets demonstrate the effectiveness of MStein over baselines. More quantitative analyses show the robustness against perturbations and training efficiency in batch size. Finally, improvements analysis indicates better representations of popular users or items with significant uncertainty. The source code is at https://github.com/zfan20/MStein.

Ziwei Fan, Zhiwei Liu, Chen Wang et al. · salesforce

Sequential Recommendation (SR) models user dynamics and predicts the next preferred items based on the user history. Existing SR methods model the 'was interacted before' item-item transitions observed in sequences, which can be viewed as an item relationship. However, there are multiple auxiliary item relationships, e.g., items from similar brands and with similar contents in real-world scenarios. Auxiliary item relationships describe item-item affinities in multiple different semantics and alleviate the long-lasting cold start problem in the recommendation. However, it remains a significant challenge to model auxiliary item relationships in SR. To simultaneously model high-order item-item transitions in sequences and auxiliary item relationships, we propose a Multi-relational Transformer capable of modeling auxiliary item relationships for SR (MT4SR). Specifically, we propose a novel self-attention module, which incorporates arbitrary item relationships and weights item relationships accordingly. Second, we regularize intra-sequence item relationships with a novel regularization module to supervise attentions computations. Third, for inter-sequence item relationship pairs, we introduce a novel inter-sequence related items modeling module. Finally, we conduct experiments on four benchmark datasets and demonstrate the effectiveness of MT4SR over state-of-the-art methods and the improvements on the cold start problem. The code is available at https://github.com/zfan20/MT4SR.

Ziwei Fan, Ke Xu, Zhang Dong et al.

Graph collaborative filtering (GCF) is a popular technique for capturing high-order collaborative signals in recommendation systems. However, GCF's bipartite adjacency matrix, which defines the neighbors being aggregated based on user-item interactions, can be noisy for users/items with abundant interactions and insufficient for users/items with scarce interactions. Additionally, the adjacency matrix ignores user-user and item-item correlations, which can limit the scope of beneficial neighbors being aggregated. In this work, we propose a new graph adjacency matrix that incorporates user-user and item-item correlations, as well as a properly designed user-item interaction matrix that balances the number of interactions across all users. To achieve this, we pre-train a graph-based recommendation method to obtain users/items embeddings, and then enhance the user-item interaction matrix via top-K sampling. We also augment the symmetric user-user and item-item correlation components to the adjacency matrix. Our experiments demonstrate that the enhanced user-item interaction matrix with improved neighbors and lower density leads to significant benefits in graph-based recommendation. Moreover, we show that the inclusion of user-user and item-item correlations can improve recommendations for users with both abundant and insufficient interactions. The code is in \url{https://github.com/zfan20/GraphDA}.

Ziwei Fan, Zhiwei Liu, Hao Peng et al. · salesforce

Sequential recommendation (SR) models the dynamic user preferences and generates the next-item prediction as the affinity between the sequence and items, in a joint latent space with low dimensions (i.e., the sequence and item embedding space). Both sequence and item representations suffer from the representation degeneration issue due to the user/item long-tail distributions, where tail users/ items are indistinguishably distributed as a narrow cone in the latent space. We argue that the representation degeneration issue is the root cause of insufficient recommendation diversity in existing SR methods, impairing the user potential exploration and further worsening the echo chamber issue. In this work, we first disclose the connection between the representation degeneration and recommendation diversity, in which severer representation degeneration indicates lower recommendation diversity. We then propose a novel Singular sPectrum sMoothing regularization for Recommendation (SPMRec), which acts as a controllable surrogate to alleviate the degeneration and achieve the balance between recommendation diversity and performance. The proposed smoothing regularization alleviates the degeneration by maximizing the area under the singular value curve, which is also the diversity surrogate. We conduct experiments on four benchmark datasets to demonstrate the superiority of SPMRec, and show that the proposed singular spectrum smoothing can control the balance of recommendation performance and diversity simultaneously.

Ziwei Fan, Hao Ding, Anoop Deoras et al.

Item-to-Item (I2I) recommendation is an important function in most recommendation systems, which generates replacement or complement suggestions for a particular item based on its semantic similarities to other cataloged items. Given that subsets of items in a recommendation system might be co-interacted with by the same set of customers, graph-based models, such as graph neural networks (GNNs), provide a natural framework to combine, ingest and extract valuable insights from such high-order relational interactions between cataloged items, as well as their metadata features, as has been shown in many recent studies. However, learning GNNs effectively for I2I requires ingesting a large amount of relational data, which might not always be available, especially in new, emerging market segments. To mitigate this data bottleneck, we postulate that recommendation patterns learned from existing mature market segments (with private data) could be adapted to build effective warm-start models for emerging ones. To achieve this, we propose and investigate a personalized federated modeling framework based on GNNs to summarize, assemble and adapt recommendation patterns across market segments with heterogeneous customer behaviors into effective local models. Our key contribution is a personalized graph adaptation model that bridges the gap between recent literature on federated GNNs and (non-graph) personalized federated learning, which either does not optimize for the adaptability of the federated model or is restricted to local models with homogeneous parameterization, excluding GNNs with heterogeneous local graphs.

Ziwei Fan, Alice Wang, Zahra Nazari

Recommender systems (RS) commonly retrieve potential candidate items for users from a massive number of items by modeling user interests based on historical interactions. However, historical interaction data is highly sparse, and most items are long-tail items, which limits the representation learning for item discovery. This problem is further augmented by the discovery of novel or cold-start items. For example, after a user displays interest in bitcoin financial investment shows in the podcast space, a recommender system may want to suggest, e.g., a newly released blockchain episode from a more technical show. Episode correlations help the discovery, especially when interaction data of episodes is limited. Accordingly, we build upon the classical Two-Tower model and introduce the novel Multi-Source Augmentations using a Contrastive Learning framework (MSACL) to enhance episode embedding learning by incorporating positive episodes from numerous correlated semantics. Extensive experiments on a real-world podcast recommendation dataset from a large audio streaming platform demonstrate the effectiveness of the proposed framework for user podcast exploration and cold-start episode recommendation.

Henry Peng Zou, Gavin Heqing Yu, Ziwei Fan et al.

In e-commerce, accurately extracting product attribute values from multimodal data is crucial for improving user experience and operational efficiency of retailers. However, previous approaches to multimodal attribute value extraction often struggle with implicit attribute values embedded in images or text, rely heavily on extensive labeled data, and can easily confuse similar attribute values. To address these issues, we introduce EIVEN, a data- and parameter-efficient generative framework that pioneers the use of multimodal LLM for implicit attribute value extraction. EIVEN leverages the rich inherent knowledge of a pre-trained LLM and vision encoder to reduce reliance on labeled data. We also introduce a novel Learning-by-Comparison technique to reduce model confusion by enforcing attribute value comparison and difference identification. Additionally, we construct initial open-source datasets for multimodal implicit attribute value extraction. Our extensive experiments reveal that EIVEN significantly outperforms existing methods in extracting implicit attribute values while requiring less labeled data.

Zihao Xu, John Harvill, Ziwei Fan et al.

Large Language Models (LLMs) incur significant computational and memory costs when processing long prompts, as full self-attention scales quadratically with input length. Token compression aims to address this challenge by reducing the number of tokens representing inputs. However, existing prompt-compression approaches primarily operate in token space and overlook inefficiencies in the latent embedding space. In this paper, we propose K-Token Merging, a latent-space compression framework that merges each contiguous block of K token embeddings into a single embedding via a lightweight encoder. The compressed sequence is processed by a LoRA-adapted LLM, while generation remains in the original vocabulary. Experiments on structural reasoning (Textualized Tree), sentiment classification (Amazon Reviews), and code editing (CommitPackFT) show that K-Token Merging lies on the Pareto frontier of performance vs. compression, achieving up to 75% input length reduction with minimal performance degradation.

Ziwei Fan, Zhiwei Liu, Alice Wang et al.

Sequential recommendation models the dynamics of a user's previous behaviors in order to forecast the next item, and has drawn a lot of attention. Transformer-based approaches, which embed items as vectors and use dot-product self-attention to measure the relationship between items, demonstrate superior capabilities among existing sequential methods. However, users' real-world sequential behaviors are \textit{\textbf{uncertain}} rather than deterministic, posing a significant challenge to present techniques. We further suggest that dot-product-based approaches cannot fully capture \textit{\textbf{collaborative transitivity}}, which can be derived in item-item transitions inside sequences and is beneficial for cold start items. We further argue that BPR loss has no constraint on positive and sampled negative items, which misleads the optimization. We propose a novel \textbf{STO}chastic \textbf{S}elf-\textbf{A}ttention~(STOSA) to overcome these issues. STOSA, in particular, embeds each item as a stochastic Gaussian distribution, the covariance of which encodes the uncertainty. We devise a novel Wasserstein Self-Attention module to characterize item-item position-wise relationships in sequences, which effectively incorporates uncertainty into model training. Wasserstein attentions also enlighten the collaborative transitivity learning as it satisfies triangle inequality. Moreover, we introduce a novel regularization term to the ranking loss, which assures the dissimilarity between positive and the negative items. Extensive experiments on five real-world benchmark datasets demonstrate the superiority of the proposed model over state-of-the-art baselines, especially on cold start items. The code is available in \url{https://github.com/zfan20/STOSA}.

Yu Wang, Zhiwei Liu, Ziwei Fan et al.

In the information explosion era, recommender systems (RSs) are widely studied and applied to discover user-preferred information. A RS performs poorly when suffering from the cold-start issue, which can be alleviated if incorporating Knowledge Graphs (KGs) as side information. However, most existing works neglect the facts that node degrees in KGs are skewed and massive amount of interactions in KGs are recommendation-irrelevant. To address these problems, in this paper, we propose Differentiable Sampling on Knowledge Graph for Recommendation with Relational GNN (DSKReG) that learns the relevance distribution of connected items from KGs and samples suitable items for recommendation following this distribution. We devise a differentiable sampling strategy, which enables the selection of relevant items to be jointly optimized with the model training procedure. The experimental results demonstrate that our model outperforms state-of-the-art KG-based recommender systems. The code is available online at https://github.com/YuWang-1024/DSKReG.

Ziwei Fan, Zhiwei Liu, Lei Zheng et al.

The sequential patterns within the user interactions are pivotal for representing the user's preference and capturing latent relationships among items. The recent advancements of sequence modeling by Transformers advocate the community to devise more effective encoders for the sequential recommendation. Most existing sequential methods assume users are deterministic. However, item-item transitions might fluctuate significantly in several item aspects and exhibit randomness of user interests. This \textit{stochastic characteristics} brings up a solid demand to include uncertainties in representing sequences and items. Additionally, modeling sequences and items with uncertainties expands users' and items' interaction spaces, thus further alleviating cold-start problems. In this work, we propose a Distribution-based Transformer for Sequential Recommendation (DT4SR), which injects uncertainties into sequential modeling. We use Elliptical Gaussian distributions to describe items and sequences with uncertainty. We describe the uncertainty in items and sequences as Elliptical Gaussian distribution. And we adopt Wasserstein distance to measure the similarity between distributions. We devise two novel Trans-formers for modeling mean and covariance, which guarantees the positive-definite property of distributions. The proposed method significantly outperforms the state-of-the-art methods. The experiments on three benchmark datasets also demonstrate its effectiveness in alleviating cold-start issues. The code is available inhttps://github.com/DyGRec/DT4SR.

Zhiwei Liu, Ziwei Fan, Yu Wang et al.

Sequential Recommendation characterizes the evolving patterns by modeling item sequences chronologically. The essential target of it is to capture the item transition correlations. The recent developments of transformer inspire the community to design effective sequence encoders, \textit{e.g.,} SASRec and BERT4Rec. However, we observe that these transformer-based models suffer from the cold-start issue, \textit{i.e.,} performing poorly for short sequences. Therefore, we propose to augment short sequences while still preserving original sequential correlations. We introduce a new framework for \textbf{A}ugmenting \textbf{S}equential \textbf{Re}commendation with \textbf{P}seudo-prior items~(ASReP). We firstly pre-train a transformer with sequences in a reverse direction to predict prior items. Then, we use this transformer to generate fabricated historical items at the beginning of short sequences. Finally, we fine-tune the transformer using these augmented sequences from the time order to predict the next item. Experiments on two real-world datasets verify the effectiveness of ASReP. The code is available on \url{https://github.com/DyGRec/ASReP}.

Zhiwei Liu, Xiaohan Li, Ziwei Fan et al.

The problem of basket recommendation~(BR) is to recommend a ranking list of items to the current basket. Existing methods solve this problem by assuming the items within the same basket are correlated by one semantic relation, thus optimizing the item embeddings. However, this assumption breaks when there exist multiple intents within a basket. For example, assuming a basket contains \{\textit{bread, cereal, yogurt, soap, detergent}\} where \{\textit{bread, cereal, yogurt}\} are correlated through the "breakfast" intent, while \{\textit{soap, detergent}\} are of "cleaning" intent, ignoring multiple relations among the items spoils the ability of the model to learn the embeddings. To resolve this issue, it is required to discover the intents within the basket. However, retrieving a multi-intent pattern is rather challenging, as intents are latent within the basket. Additionally, intents within the basket may also be correlated. Moreover, discovering a multi-intent pattern requires modeling high-order interactions, as the intents across different baskets are also correlated. To this end, we propose a new framework named as \textbf{M}ulti-\textbf{I}ntent \textbf{T}ranslation \textbf{G}raph \textbf{N}eural \textbf{N}etwork~({\textbf{MITGNN}}). MITGNN models $T$ intents as tail entities translated from one corresponding basket embedding via $T$ relation vectors. The relation vectors are learned through multi-head aggregators to handle user and item information. Additionally, MITGNN propagates multiple intents across our defined basket graph to learn the embeddings of users and items by aggregating neighbors. Extensive experiments on two real-world datasets prove the effectiveness of our proposed model on both transductive and inductive BR. The code is available online at https://github.com/JimLiu96/MITGNN.

Behnam Rahdari, Hao Ding, Ziwei Fan et al.

The unique capabilities of Large Language Models (LLMs), such as the natural language text generation ability, position them as strong candidates for providing explanation for recommendations. However, despite the size of the LLM, most existing models struggle to produce zero-shot explanations reliably. To address this issue, we propose a framework called Logic-Scaffolding, that combines the ideas of aspect-based explanation and chain-of-thought prompting to generate explanations through intermediate reasoning steps. In this paper, we share our experience in building the framework and present an interactive demonstration for exploring our results.

Sen Yang, Minyue Jiang, Ziwei Fan et al.

Recent advances in autonomous driving systems have shifted towards reducing reliance on high-definition maps (HDMaps) due to the huge costs of annotation and maintenance. Instead, researchers are focusing on online vectorized HDMap construction using on-board sensors. However, sensor-only approaches still face challenges in long-range perception due to the restricted views imposed by the mounting angles of onboard cameras, just as human drivers also rely on bird's-eye-view navigation maps for a comprehensive understanding of road structures. To address these issues, we propose to train the perception model to "see" standard definition maps (SDMaps). We encode SDMap elements into neural spatial map representations and instance tokens, and then incorporate such complementary features as prior information to improve the bird's eye view (BEV) feature for lane geometry and topology decoding. Based on the lane segment representation framework, the model simultaneously predicts lanes, centrelines and their topology. To further enhance the ability of geometry prediction and topology reasoning, we also use a topology-guided decoder to refine the predictions by exploiting the mutual relationships between topological and geometric features. We perform extensive experiments on OpenLane-V2 datasets to validate the proposed method. The results show that our model outperforms state-of-the-art methods by a large margin, with gains of +6.7 and +9.1 on the mAP and topology metrics. Our analysis also reveals that models trained with SDMap noise augmentation exhibit enhanced robustness.

John Harvill, Ziwei Fan, Hao Wang et al.

Existing work on prompt compression for Large Language Models (LLM) focuses on lossy methods that try to maximize the retention of semantic information that is relevant to downstream tasks while significantly reducing the sequence length. In this paper, we introduce a task-agnostic lossless compression technique similar to LZ77 that makes it possible to reduce the input token sequence length on average by 27\% and 18\% for the two evaluation tasks explored here. Given that we use transformer-based LLMs, this equates to 47\% and 33\% less encoding computation, respectively, due to the quadratic nature of attention. The token sequence transformation is trivial to reverse and highlights that no semantic information is lost in the process. We evaluate our proposed approach on two tasks that require strict preservation of semantics/syntax and demonstrate that existing lossy compression methods perform poorly in this setting. We find that our lossless compression technique produces only a small gap in performance compared to using the uncompressed input and posit that larger models and an expanded computing budget would likely erase the gap entirely.

Ziwei Fan, Zhiwei Liu, Shelby Heinecke et al.

Existing recommender systems face difficulties with zero-shot items, i.e. items that have no historical interactions with users during the training stage. Though recent works extract universal item representation via pre-trained language models (PLMs), they ignore the crucial item relationships. This paper presents a novel paradigm for the Zero-Shot Item-based Recommendation (ZSIR) task, which pre-trains a model on product knowledge graph (PKG) to refine the item features from PLMs. We identify three challenges for pre-training PKG, which are multi-type relations in PKG, semantic divergence between item generic information and relations and domain discrepancy from PKG to downstream ZSIR task. We address the challenges by proposing four pre-training tasks and novel task-oriented adaptation (ToA) layers. Moreover, this paper discusses how to fine-tune the model on new recommendation task such that the ToA layers are adapted to ZSIR task. Comprehensive experiments on 18 markets dataset are conducted to verify the effectiveness of the proposed model in both knowledge prediction and ZSIR task.

Liangwei Yang, Zhiwei Liu, Yu Wang et al.

Because of the large number of online games available nowadays, online game recommender systems are necessary for users and online game platforms. The former can discover more potential online games of their interests, and the latter can attract users to dwell longer in the platform. This paper investigates the characteristics of user behaviors with respect to the online games on the Steam platform. Based on the observations, we argue that a satisfying recommender system for online games is able to characterize: personalization, game contextualization and social connection. However, simultaneously solving all is rather challenging for game recommendation. Firstly, personalization for game recommendation requires the incorporation of the dwelling time of engaged games, which are ignored in existing methods. Secondly, game contextualization should reflect the complex and high-order properties of those relations. Last but not least, it is problematic to use social connections directly for game recommendations due to the massive noise within social connections. To this end, we propose a Social-aware Contextualized Graph Neural Recommender System (SCGRec), which harnesses three perspectives to improve game recommendation. We conduct a comprehensive analysis of users' online game behaviors, which motivates the necessity of handling those three characteristics in the online game recommendation.

Zhiwei Liu, Liangwei Yang, Ziwei Fan et al.

Recommender systems have become prosperous nowadays, designed to predict users' potential interests in items by learning embeddings. Recent developments of the Graph Neural Networks~(GNNs) also provide recommender systems with powerful backbones to learn embeddings from a user-item graph. However, only leveraging the user-item interactions suffers from the cold-start issue due to the difficulty in data collection. Hence, current endeavors propose fusing social information with user-item interactions to alleviate it, which is the social recommendation problem. Existing work employs GNNs to aggregate both social links and user-item interactions simultaneously. However, they all require centralized storage of the social links and item interactions of users, which leads to privacy concerns. Additionally, according to strict privacy protection under General Data Protection Regulation, centralized data storage may not be feasible in the future, urging a decentralized framework of social recommendation. To this end, we devise a novel framework \textbf{Fe}drated \textbf{So}cial recommendation with \textbf{G}raph neural network (FeSoG). Firstly, FeSoG adopts relational attention and aggregation to handle heterogeneity. Secondly, FeSoG infers user embeddings using local data to retain personalization. Last but not least, the proposed model employs pseudo-labeling techniques with item sampling to protect the privacy and enhance training. Extensive experiments on three real-world datasets justify the effectiveness of FeSoG in completing social recommendation and privacy protection. We are the first work proposing a federated learning framework for social recommendation to the best of our knowledge.

Xiang Bai, Hanchen Wang, Liya Ma et al.

Artificial intelligence (AI) provides a promising substitution for streamlining COVID-19 diagnoses. However, concerns surrounding security and trustworthiness impede the collection of large-scale representative medical data, posing a considerable challenge for training a well-generalised model in clinical practices. To address this, we launch the Unified CT-COVID AI Diagnostic Initiative (UCADI), where the AI model can be distributedly trained and independently executed at each host institution under a federated learning framework (FL) without data sharing. Here we show that our FL model outperformed all the local models by a large yield (test sensitivity /specificity in China: 0.973/0.951, in the UK: 0.730/0.942), achieving comparable performance with a panel of professional radiologists. We further evaluated the model on the hold-out (collected from another two hospitals leaving out the FL) and heterogeneous (acquired with contrast materials) data, provided visual explanations for decisions made by the model, and analysed the trade-offs between the model performance and the communication costs in the federated training process. Our study is based on 9,573 chest computed tomography scans (CTs) from 3,336 patients collected from 23 hospitals located in China and the UK. Collectively, our work advanced the prospects of utilising federated learning for privacy-preserving AI in digital health.

Ziwei Fan, Zhiwei Liu, Jiawei Zhang et al.

In order to model the evolution of user preference, we should learn user/item embeddings based on time-ordered item purchasing sequences, which is defined as Sequential Recommendation (SR) problem. Existing methods leverage sequential patterns to model item transitions. However, most of them ignore crucial temporal collaborative signals, which are latent in evolving user-item interactions and coexist with sequential patterns. Therefore, we propose to unify sequential patterns and temporal collaborative signals to improve the quality of recommendation, which is rather challenging. Firstly, it is hard to simultaneously encode sequential patterns and collaborative signals. Secondly, it is non-trivial to express the temporal effects of collaborative signals. Hence, we design a new framework Temporal Graph Sequential Recommender (TGSRec) upon our defined continuous-time bi-partite graph. We propose a novel Temporal Collaborative Trans-former (TCT) layer in TGSRec, which advances the self-attention mechanism by adopting a novel collaborative attention. TCT layer can simultaneously capture collaborative signals from both users and items, as well as considering temporal dynamics inside sequential patterns. We propagate the information learned fromTCTlayerover the temporal graph to unify sequential patterns and temporal collaborative signals. Empirical results on five datasets show that TGSRec significantly outperforms other baselines, in average up to 22.5% and 22.1%absolute improvements in Recall@10and MRR, respectively.

Ziwei Fan, Evan Burgun, Zhiyun Ren et al.

Due to the rapid growth of information available about individual patients, most physicians suffer from information overload when they review patient information in health information technology systems. In this manuscript, we present a novel hybrid dynamic and multi-collaborative filtering method to improve information retrieval from electronic health records. This method recommends relevant information from electronic health records for physicians during patient visits. It models information search dynamics using a Markov model. It also leverages the key idea of collaborative filtering, originating from Recommender Systems, to prioritize information based on various similarities among physicians, patients and information items. We tested this new method using real electronic health record data from the Indiana Network for Patient Care. Our experimental results demonstrated that for 46.7% of testing cases, this new method is able to correctly prioritize relevant information among top-5 recommendations that physicians are truly interested in.