Federated recommendation (FR) facilitates collaborative training by aggregating local models from massive devices, enabling client-specific personalization while ensuring privacy. However, we empirically and theoretically demonstrate that server-side aggregation can undermine client-side personalization, leading to suboptimal performance, i.e., the aggregation bottleneck. This issue stems from the inherent heterogeneity across numerous clients in FR, which drives the global model to deviate from local optima. To this end, we propose FedEM, which elastically merges the global and local models to compensate for impaired personalization. Unlike existing personalized federated recommendation (pFR) methods, FedEM (1) investigates the aggregation bottleneck in FR through theoretical insights, rather than relying on heuristic analysis; (2) leverages off-the-shelf local models rather than designing additional mechanisms to boost personalization. Extensive experiments demonstrate that our method preserves client personalization during collaborative training, outperforming state-of-the-art baselines.

Federated recommendations (FRs), facilitating multiple local clients to collectively learn a global model without disclosing user private data, have emerged as a prevalent on-device service. In conventional FRs, a dominant paradigm is to utilize discrete identities to represent clients and items, which are then mapped to domain-specific embeddings to participate in model training. Despite considerable performance, we reveal three inherent limitations that can not be ignored in federated settings, i.e., non-transferability across domains, ineffectiveness in cold-start settings, and potential privacy violations during federated training. To this end, we propose a transferable federated recommendation model, TransFR, which delicately incorporates the general capabilities empowered by pre-trained models and the personalized abilities by fine-tuning local private data. Specifically, it first learns domain-agnostic representations of items by exploiting pre-trained models with public textual corpora. To tailor for FR tasks, we further introduce efficient federated adapter-tuning and post-adaptation personalization, which facilitate personalized adapters for each client by fitting local private data. We theoretically prove the advantages of incorporating adapter tuning in FRs regarding both effectiveness and privacy. Through extensive experiments, we show that our TransFR surpasses state-of-the-art FRs on transferability.

Federated recommendation aims to collect global knowledge by aggregating local models from massive devices, to provide recommendations while ensuring privacy. Current methods mainly leverage aggregation functions invented by federated vision community to aggregate parameters from similar clients, e.g., clustering aggregation. Despite considerable performance, we argue that it is suboptimal to apply them to federated recommendation directly. This is mainly reflected in the disparate model structures. Different from structured parameters like convolutional neural networks in federated vision, federated recommender models usually distinguish itself by employing one-to-one item embedding table. Such a discrepancy induces the challenging embedding skew issue, which continually updates the trained embeddings but ignores the non-trained ones during aggregation, thus failing to predict future items accurately. To this end, we propose a personalized Federated recommendation model with Composite Aggregation (FedCA), which not only aggregates similar clients to enhance trained embeddings but also aggregates complementary clients to update non-trained embeddings. Besides, we formulate the overall learning process into a unified optimization algorithm to jointly learn the similarity and complementarity. Extensive experiments on several real-world datasets substantiate the effectiveness of our proposed model. Our code is available at https://github.com/hongleizhang/FedCA .

Federated machine unlearning service is an emerging Machine-Learning-as-a-Service (MLaaS) paradigm which supports the request of removing or forgetting the influence of a specific group of data from federated learning services. While current federated unlearning methods work well in removing instances on individual clients, they encounter challenges in addressing multiple forms of unlearning requirements, such as heterogeneous learning models and diverse unlearning contents. In this paper, we propose novel federated few-shot learning and unlearning models inspired by distillation learning. Firstly, we propose a federated metric learning model where only prototypes are transmitted between the server and clients with limited samples. The training data with different input dimensions on local clients are transformed into abstract prototypes with the same length, enabling collaborative training of heterogeneous models among clients. Then, we propose an efficient federated metric unlearning method, where the temporarily stored prototypes are utilized as teacher knowledge to guide and accelerate the retraining process for each unlearning scenario. The complexity of the federated metric unlearning method is analyzed to show the computation time and communication efficiency. Experimental results demonstrate that our approach outperforms baseline methods in terms of accuracy while effectively removing various requested unlearning contents.

Current collaborative perception methods often rely on fully annotated datasets, which can be expensive to obtain in practical situations. To reduce annotation costs, some works adopt sparsely supervised learning techniques and generate pseudo labels for the missing instances. However, these methods fail to achieve an optimal confidence threshold that harmonizes the quality and quantity of pseudo labels. To address this issue, we propose an end-to-end Collaborative perception Dual Teacher-Student framework (CoDTS), which employs adaptive complementary learning to produce both high-quality and high-quantity pseudo labels. Specifically, the Main Foreground Mining (MFM) module generates high-quality pseudo labels based on the prediction of the static teacher. Subsequently, the Supplement Foreground Mining (SFM) module ensures a balance between the quality and quantity of pseudo labels by adaptively identifying missing instances based on the prediction of the dynamic teacher. Additionally, the Neighbor Anchor Sampling (NAS) module is incorporated to enhance the representation of pseudo labels. To promote the adaptive complementary learning, we implement a staged training strategy that trains the student and dynamic teacher in a mutually beneficial manner. Extensive experiments demonstrate that the CoDTS effectively ensures an optimal balance of pseudo labels in both quality and quantity, establishing a new state-of-the-art in sparsely supervised collaborative perception.

Federated recommender systems (FedRecSys) have emerged as a pivotal solution for privacy-aware recommendations, balancing growing demands for data security and personalized experiences. Current research efforts predominantly concentrate on adapting traditional recommendation architectures to federated environments, optimizing communication efficiency, and mitigating security vulnerabilities. However, user personalization modeling, which is essential for capturing heterogeneous preferences in this decentralized and non-IID data setting, remains underexplored. This survey addresses this gap by systematically exploring personalization in FedRecSys, charting its evolution from centralized paradigms to federated-specific innovations. We establish a foundational definition of personalization in a federated setting, emphasizing personalized models as a critical solution for capturing fine-grained user preferences. The work critically examines the technical hurdles of building personalized FedRecSys and synthesizes promising methodologies to meet these challenges. As the first consolidated study in this domain, this survey serves as both a technical reference and a catalyst for advancing personalized FedRecSys research.

Federated recommender systems (FRSs), with their improved privacy-preserving advantages to jointly train recommendation models from numerous devices while keeping user data distributed, have been widely explored in modern recommender systems (RSs). However, conventional FRSs require transmitting the entire model between the server and clients, which brings a huge carbon footprint for cost-conscious cross-device learning tasks. While several efforts have been dedicated to improving the efficiency of FRSs, it's suboptimal to treat the whole model as the objective of compact design. Besides, current research fails to handle the out-of-vocabulary (OOV) issue in real-world FRSs, where the items only occasionally appear in the testing phase but were not observed during the training process, which is another practical challenge and has not been well studied yet. To this end, we propose a privacy-enhanced federated recommendation framework with shared hash embedding, PrivFR, in cross-device settings, which is an efficient representation mechanism specialized for the embedding parameters without compromising the model capability. Specifically, it represents items in a resource-efficient way by delicately utilizing shared hash embedding and multiple hash functions. As such, it just maintains a small shared pool of hash embedding in local clients, rather than fitting all embedding vectors for each item, which can exactly achieve the dual advantages of conserving resources and handling the OOV issue. What's more, we prove that this mechanism can protect the data privacy of local clients from a theoretical perspective. Extensive experiments show that our method not only effectively reduces storage and communication overheads, but also outperforms state-of-the-art FRSs.

Recommendation systems are pivotal in improving user experience on various digital platforms. However, observational training data in recommendation systems introduce selection bias, which leads to a distributional discrepancy between training data and real-world scenarios, resulting in suboptimal performance. Current causal debiasing methods such as inverse propensity score and doubly robust rely on accurately estimated propensity scores, typically optimized through negative log-likelihood (NLL) minimization. However, recent studies have highlighted the limitations of this approach, as perfect NLL minimization may not adequately correct for selection bias. To address this issue, we propose Wasserstein Balancing Metric (WBM), a novel metric that measures and enhances the balancing capacity of propensity scores in causal debiasing methods by minimizing the Wasserstein discrepancy between reweighted populations. On the basis, we introduce IPS-WBM and DR-WBM, incorporating WBM as a regularizer in standard inverse propensity score and doubly robust estimators, which enhances causal balancing capacity without introducing additional bias. Extensive experiments on three real-world recommendation datasets demonstrate that our methods improve the causal balancing capability of learned propensities and enhance debiasing performance.

In database of recommender systems, users' ratings for most items are usually missing, resulting in selection bias when users selectively choose items to rate. To address this problem, propensity-based methods, e.g., inverse propensity scoring and doubly robust, have been widely studied and applied to missing rating prediction and post-click conversion rate prediction tasks. However, have we completely eliminated the selection bias? Under what missing data mechanism can previous studies completely eliminate the selection bias and lead to unbiased learning? In this paper, following the previous literature on statistics, we first formally define three missing data mechanisms, i.e., missing completely at random (MCAR), missing at random (MAR), and missing not at random (MNAR), and discuss the widespread prevalence of MNAR in recommender systems. Next, we theoretically reveal that the unbiasedness of previous propensity-based debiasing methods is valid only when data are MCAR or MAR, while it leads to biased predictions when data are MNAR. To tackle this research gap, we propose to disentangle user and item embeddings into the primary latent vector for rating prediction and the auxiliary latent vector for missing mechanism modeling. We prove the identifiablility results, and show that the proposed method can achieve unbiased learning under MNAR with imposed constraints. Extensive experiments are conducted on a semi-synthetic dataset and three real-world datasets, validating the effectiveness of our proposed method.

Collaborative 3D object detection, with its improved interaction advantage among multiple agents, has been widely explored in autonomous driving. However, existing collaborative 3D object detectors in a fully supervised paradigm heavily rely on large-scale annotated 3D bounding boxes, which is labor-intensive and time-consuming. To tackle this issue, we propose a sparsely supervised collaborative 3D object detection framework SSC3OD, which only requires each agent to randomly label one object in the scene. Specifically, this model consists of two novel components, i.e., the pillar-based masked autoencoder (Pillar-MAE) and the instance mining module. The Pillar-MAE module aims to reason over high-level semantics in a self-supervised manner, and the instance mining module generates high-quality pseudo labels for collaborative detectors online. By introducing these simple yet effective mechanisms, the proposed SSC3OD can alleviate the adverse impacts of incomplete annotations. We generate sparse labels based on collaborative perception datasets to evaluate our method. Extensive experiments on three large-scale datasets reveal that our proposed SSC3OD can effectively improve the performance of sparsely supervised collaborative 3D object detectors.

Federated recommender system (FRS), which enables many local devices to train a shared model jointly without transmitting local raw data, has become a prevalent recommendation paradigm with privacy-preserving advantages. However, previous work on FRS performs similarity search via inner product in continuous embedding space, which causes an efficiency bottleneck when the scale of items is extremely large. We argue that such a scheme in federated settings ignores the limited capacities in resource-constrained user devices ( i.e. , storage space, computational overhead, and communication bandwidth), and makes it harder to be deployed in large-scale recommender systems. Besides, it has been shown that transmitting local gradients in real-valued form between server and clients may leak users’ private information. To this end, we propose a lightweight federated recommendation framework with privacy-preserving matrix factorization, LightFR , that is able to generate high-quality binary codes by exploiting learning to hash technique under federated settings, and thus enjoys both fast online inference and economic memory consumption. Moreover, we devise an efficient federated discrete optimization algorithm to collaboratively train model parameters between the server and clients, which can effectively prevent real-valued gradient attacks from malicious parties. Through extensive experiments on four real-world datasets, we show that our LightFR model outperforms several state-of-the-art FRS methods in terms of recommendation accuracy, inference efficiency and data privacy.

To address the data sparsity problem faced by recommender systems, social network among users is often utilized to complement rating data for improving the recommendation performance. One of current trends is to combine the idea of matrix factorization (MF) for predicting ratings with the idea of graph embedding (GE) for analyzing social network towards recommendation tasks. Despite enjoying many advantages, the existing integrated models have two critical limitations. First, such models are designed to work with either explicit or implicit social network, but little is known in taking both into account. Second, the users’ embeddings learned by GE are fed to the downstream MF, but not reverse, which is sub-optimal because rating information is not considered for learning the users’ embeddings. In this paper, we propose a novel social recommendation algorithm which exploits both explicit and implicit social networks towards the task of rating prediction. In Particular, we seamlessly integrate MF model and GE model within a unified optimization framework, in which MF and GE tasks can be reinforced each other during the learning process. Our encouraging experimental results on three real-world benchmarks validate the superiority of the proposed approach to state-of-the-art social recommendation methods.