Delving into Differentially Private Transformer

Deep learning with differential privacy (DP) has garnered significant attention over the past years, leading to the development of numerous methods aimed at enhancing model accuracy and training efficiency. This paper delves into the problem of training Transformer models with differential privacy. Our treatment is modular: the logic is to `reduce' the problem of training DP Transformer to the more basic problem of training DP vanilla neural nets. The latter is better understood and amenable to many model-agnostic methods. Such `reduction' is done by first identifying the hardness unique to DP Transformer training: the attention distraction phenomenon and a lack of compatibility with existing techniques for efficient gradient clipping. To deal with these two issues, we propose the Re-Attention Mechanism and Phantom Clipping, respectively. We believe that our work not only casts new light on training DP Transformers but also promotes a modular treatment to advance research in the field of differentially private deep learning.

BMLP: Behavior-aware MLP for Heterogeneous Sequential Recommendation

In real recommendation scenarios, users often have different types of behaviors, such as clicking and buying. Existing research methods show that it is possible to capture the heterogeneous interests of users through different types of behaviors. However, most multi-behavior approaches have limitations in learning the relationship between different behaviors. In this paper, we propose a novel multilayer perceptron (MLP)-based heterogeneous sequential recommendation method, namely behavior-aware multilayer perceptron (BMLP). Specifically, it has two main modules, including a heterogeneous interest perception (HIP) module, which models behaviors at multiple granularities through behavior types and transition relationships, and a purchase intent perception (PIP) module, which adaptively fuses subsequences of auxiliary behaviors to capture users' purchase intent. Compared with mainstream sequence models, MLP is competitive in terms of accuracy and has unique advantages in simplicity and efficiency. Extensive experiments show that BMLP achieves significant improvement over state-of-the-art algorithms on four public datasets. In addition, its pure MLP architecture leads to a linear time complexity.

Privacy-Preserving Cross-Domain Sequential Recommendation

Cross-domain sequential recommendation is an important development direction of recommender systems. It combines the characteristics of sequential recommender systems and cross-domain recommender systems, which can capture the dynamic preferences of users and alleviate the problem of cold-start users. However, in recent years, people pay more and more attention to their privacy. They do not want other people to know what they just bought, what videos they just watched, and where they just came from. How to protect the users' privacy has become an urgent problem to be solved. In this paper, we propose a novel privacy-preserving cross-domain sequential recommender system (PriCDSR), which can provide users with recommendation services while preserving their privacy at the same time. Specifically, we define a new differential privacy on the data, taking into account both the ID information and the order information. Then, we design a random mechanism that satisfies this differential privacy and provide its theoretical proof. Our PriCDSR is a non-invasive method that can adopt any cross-domain sequential recommender system as a base model without any modification to it. To the best of our knowledge, our PriCDSR is the first work to investigate privacy issues in cross-domain sequential recommender systems. We conduct experiments on three domains, and the results demonstrate that our PriCDSR, despite introducing noise, still outperforms recommender systems that only use data from a single domain.

A Survey on Cross-Domain Sequential Recommendation

Cross-domain sequential recommendation (CDSR) shifts the modeling of user preferences from flat to stereoscopic by integrating and learning interaction information from multiple domains at different granularities (ranging from inter-sequence to intra-sequence and from single-domain to cross-domain). In this survey, we first define the CDSR problem using a four-dimensional tensor and then analyze its multi-type input representations under multidirectional dimensionality reductions. Following that, we provide a systematic overview from both macro and micro views. From a macro view, we abstract the multi-level fusion structures of various models across domains and discuss their bridges for fusion. From a micro view, focusing on the existing models, we specifically discuss the basic technologies and then explain the auxiliary learning technologies. Finally, we exhibit the available public datasets and the representative experimental results as well as provide some insights into future directions for research in CDSR.

A Survey on Multi-Behavior Sequential Recommendation

Recommender systems is set up to address the issue of information overload in traditional information retrieval systems, which is focused on recommending information that is of most interest to users from massive information. Generally, there is a sequential nature and heterogeneity to the behavior of a person interacting with a system, leading to the proposal of multi-behavior sequential recommendation (MBSR). MBSR is a relatively new and worthy direction for in-depth research, which can achieve state-of-the-art recommendation through suitable modeling, and some related works have been proposed. This survey aims to shed light on the MBSR problem. Firstly, we introduce MBSR in detail, including its problem definition, application scenarios and challenges faced. Secondly, we detail the classification of MBSR, including neighborhood-based methods, matrix factorization-based methods and deep learning-based methods, where we further classify the deep learning-based methods into different learning architectures based on RNN, GNN, Transformer, and generic architectures as well as architectures that integrate hybrid techniques. In each method, we present related works based on the data perspective and the modeling perspective, as well as analyze the strengths, weaknesses and features of these works. Finally, we discuss some promising future research directions to address the challenges and improve the current status of MBSR.

GNN4FR: A Lossless GNN-based Federated Recommendation Framework

Graph neural networks (GNNs) have gained wide popularity in recommender systems due to their capability to capture higher-order structure information among the nodes of users and items. However, these methods need to collect personal interaction data between a user and the corresponding items and then model them in a central server, which would break the privacy laws such as GDPR. So far, no existing work can construct a global graph without leaking each user's private interaction data (i.e., his or her subgraph). In this paper, we are the first to design a novel lossless federated recommendation framework based on GNN, which achieves full-graph training with complete high-order structure information, enabling the training process to be equivalent to the corresponding un-federated counterpart. In addition, we use LightGCN to instantiate an example of our framework and show its equivalence.

DPFormer: Learning Differentially Private Transformer on Long-Tailed Data

The Transformer has emerged as a versatile and effective architecture with broad applications. However, it still remains an open problem how to efficiently train a Transformer model of high utility with differential privacy guarantees. In this paper, we identify two key challenges in learning differentially private Transformers, i.e., heavy computation overhead due to per-sample gradient clipping and unintentional attention distraction within the attention mechanism. In response, we propose DPFormer, equipped with Phantom Clipping and Re-Attention Mechanism, to address these challenges. Our theoretical analysis shows that DPFormer can reduce computational costs during gradient clipping and effectively mitigate attention distraction (which could obstruct the training process and lead to a significant performance drop, especially in the presence of long-tailed data). Such analysis is further corroborated by empirical results on two real-world datasets, demonstrating the efficiency and effectiveness of the proposed DPFormer.

Bounding System-Induced Biases in Recommender Systems with A Randomized Dataset

Debiased recommendation with a randomized dataset has shown very promising results in mitigating the system-induced biases. However, it still lacks more theoretical insights or an ideal optimization objective function compared with the other more well studied route without a randomized dataset. To bridge this gap, we study the debiasing problem from a new perspective and propose to directly minimize the upper bound of an ideal objective function, which facilitates a better potential solution to the system-induced biases. Firstly, we formulate a new ideal optimization objective function with a randomized dataset. Secondly, according to the prior constraints that an adopted loss function may satisfy, we derive two different upper bounds of the objective function, i.e., a generalization error bound with the triangle inequality and a generalization error bound with the separability. Thirdly, we show that most existing related methods can be regarded as the insufficient optimization of these two upper bounds. Fourthly, we propose a novel method called debiasing approximate upper bound with a randomized dataset (DUB), which achieves a more sufficient optimization of these upper bounds. Finally, we conduct extensive experiments on a public dataset and a real product dataset to verify the effectiveness of our DUB.

Self-Sampling Training and Evaluation for the Accuracy-Bias Tradeoff in Recommendation

Research on debiased recommendation has shown promising results. However, some issues still need to be handled for its application in industrial recommendation. For example, most of the existing methods require some specific data, architectures and training methods. In this paper, we first argue through an online study that arbitrarily removing all the biases in industrial recommendation may not consistently yield a desired performance improvement. For the situation that a randomized dataset is not available, we propose a novel self-sampling training and evaluation (SSTE) framework to achieve the accuracy-bias tradeoff in recommendation, i.e., eliminate the harmful biases and preserve the beneficial ones. Specifically, SSTE uses a self-sampling module to generate some subsets with different degrees of bias from the original training and validation data. A self-training module infers the beneficial biases and learns better tradeoff based on these subsets, and a self-evaluation module aims to use these subsets to construct more plausible references to reflect the optimized model. Finally, we conduct extensive offline experiments on two datasets to verify the effectiveness of our SSTE. Moreover, we deploy our SSTE in homepage recommendation of a famous financial management product called Tencent Licaitong, and find very promising results in an online A/B test.

DIWIFT: Discovering Instance-wise Influential Features for Tabular Data

Tabular data is one of the most common data storage formats in business applications, ranging from retail, bank and E-commerce. These applications rely heavily on machine learning models to achieve business success. One of the critical problems in learning tabular data is to distinguish influential features from all the predetermined features. Global feature selection has been well-studied for quite some time, assuming that all instances have the same influential feature subsets. However, different instances rely on different feature subsets in practice, which also gives rise to that instance-wise feature selection receiving increasing attention in recent studies. In this paper, we first propose a novel method for discovering instance-wise influential features for tabular data (DIWIFT), the core of which is to introduce the influence function to measure the importance of an instance-wise feature. DIWIFT is capable of automatically discovering influential feature subsets of different sizes in different instances, which is different from global feature selection that considers all instances with the same influential feature subset. On the other hand, different from the previous instance-wise feature selection, DIWIFT minimizes the validation loss on the validation set and is thus more robust to the distribution shift existing in the training dataset and test dataset, which is important in tabular data. Finally, we conduct extensive experiments on both synthetic and real-world datasets to validate the effectiveness of our DIWIFT, compared it with baseline methods. Moreover, we also demonstrate the robustness of our method via some ablation experiments.