Deep Learning for Sequential Recommendation

Sequential recommender systems, as a specialized branch of recommender systems that can capture users’ dynamic preferences for more accurate and timely recommendations, have made significant progress. Recently, due to increasing concerns about user data privacy, some researchers have implemented federated learning for sequential recommendation, a.k.a., Federated Sequential Recommender Systems (FedSeqRecs), in which a public sequential recommender model is shared and frequently transmitted between a central server and clients to achieve collaborative learning. Although these solutions mitigate user privacy to some extent, they present two significant limitations that affect their practical usability: (1) They require a globally shared sequential recommendation model. However, in real-world scenarios, the recommendation model constitutes a critical intellectual property for platform and service providers. Therefore, service providers may be reluctant to disclose their meticulously developed models. (2) The communication costs are high as they correlate with the number of model parameters. This becomes particularly problematic as the current FedSeqRec will be inapplicable when sequential recommendation marches into a large language model era. To overcome the above challenges, this article proposes a parameter transmission-free federated sequential recommendation framework (PTF-FSR), which ensures both model and data privacy protection to meet the privacy needs of service providers and system users alike. Furthermore, since PTF-FSR only transmits prediction results under privacy protection, which are independent of model sizes, this new federated learning architecture can accommodate more complex and larger sequential recommendation models. Extensive experiments conducted on three widely used recommendation datasets, employing various sequential recommendation models from both ID-based and ID-free paradigms, demonstrate the effectiveness and generalization capability of our proposed framework. To facilitate future research in this direction, we release our code at https://github.com/hi-weiyuan/PTF-FSR .

In contrast to traditional recommender systems which usually pay attention to users' general and long-term preferences, sequential recommendation (SR) can model users' dynamic intents based on their behaviour sequences and suggest the next item(s) to them. However, most of existing sequential models learn the ranking score of an item only based on its relevance property, and the personalized user demands in terms of different learning objectives, such as diversity, tail novelty or recency, which have been played essential roles in multi-objective recommendation (MOR), are often neglected in SR. In this paper, we first discuss the importance of considering multiple different objectives within a learning model for recommender system. Next, to capture users' objective-level preferences by utilizing interactive information in the sequential context, we propose a novel Dynamic Multi-objective Recommendation (DMORec) framework with interactive evolution for SR. In particular, DMORec formulates a dynamic multi-objective optimization task to simultaneously optimize more than two varying objectives in an interactive recommendation process. Moreover, to resolve this optimization task in SR, we develop an evolutionary algorithm with supervised learning approach to obtain the Pareto-optimal solutions of the formulated problem. Comprehensive experiments on four real-world datasets demonstrate the effectiveness of the proposed DMORec for dynamic multi-objective recommendation in sequential recommender systems.

Sequential recommendation is a popular task in academic research and close to real-world application scenarios, where the goal is to predict the next action(s) of the user based on his/her previous sequence of actions. In the training process of recommender systems, the loss function plays an essential role in guiding the optimization of recommendation models to generate accurate suggestions for users. However, most existing sequential recommendation techniques focus on designing algorithms or neural network architectures, and few efforts have been made to tailor loss functions that fit naturally into the practical application scenario of sequential recommender systems. Ranking-based losses, such as cross-entropy and Bayesian Personalized Ranking (BPR) are widely used in the sequential recommendation area. We argue that such objective functions suffer from two inherent drawbacks: i) the dependencies among elements of a sequence are overlooked in these loss formulations; ii) instead of balancing accuracy (quality) and diversity, only generating accurate results has been over emphasized. We therefore propose two new loss functions based on the Determinantal Point Process (DPP) likelihood, that can be adaptively applied to estimate the subsequent item or items. The DPP-distributed item set captures natural dependencies among temporal actions, and a quality vs. diversity decomposition of the DPP kernel pushes us to go beyond accuracy-oriented loss functions. Experimental results using the proposed loss functions on three real-world datasets show marked improvements over state-of-the-art sequential recommendation methods in both quality and diversity metrics.

Sequential recommendation has been widely studied in the recommendation domain since it can capture users’ temporal preferences and provide more accurate and timely recommendations. To address user privacy concerns, the combination of federated learning and sequential recommender systems (FedSeqRec) has gained growing attention. Unfortunately, the performance of FedSeqRec is still unsatisfactory because the models used in FedSeqRec have to be lightweight to accommodate communication bandwidth and clients’ on-device computational resource constraints. Recently, large language models (LLMs) have exhibited strong transferable and generalized language understanding abilities and therefore, in the NLP area, many downstream tasks now utilize LLMs as a service to achieve superior performance without constructing complex models. Inspired by this successful practice, we propose a generic FedSeqRec framework, FELLAS, which aims to enhance FedSeqRec by utilizing LLMs as an external service. Specifically, FELLAS employs an LLM server to provide both item-level and sequence-level representation assistance. The item-level representation service is queried by the central server to enrich the original ID-based item embedding with textual information, while the sequence-level representation service is accessed by each client. However, invoking the sequence-level representation service requires clients to send sequences to the external LLM server. To safeguard privacy, we implement \(d_{\mathcal{X}}\) -privacy satisfied sequence perturbation, which protects clients’ sensitive data with guarantees. Additionally, a contrastive learning-based method is designed to transfer knowledge from the noisy sequence representation to clients’ sequential recommendation models. Furthermore, to empirically validate the privacy protection capability of FELLAS, we propose two interacted item inference attacks, considering the threats posed by the LLM server and the central server acting as curious-but-honest adversaries in cooperation. Extensive experiments conducted on three datasets with two widely used sequential recommendation models demonstrate the effectiveness and privacy-preserving capability of FELLAS.

Modern deep neural networks (DNNs) have greatly facilitated the development of sequential recommender systems by achieving state-of-the-art recommendation performance on various sequential recommendation tasks. Given a sequence of interacted items, existing DNN-based sequential recommenders commonly embed each item into a unique vector to support subsequent computations of the user interest. However, due to the potentially large number of items, the over-parameterised item embedding matrix of a sequential recommender has become a memory bottleneck for efficient deployment in resource-constrained environments, e.g., smartphones and other edge devices. Furthermore, we observe that the widely-used multi-head self-attention, though being effective in modelling sequential dependencies among items, heavily relies on redundant attention units to fully capture both global and local item-item transition patterns within a sequence. In this paper, we introduce a novel lightweight self-attentive network (LSAN) for sequential recommendation. To aggressively compress the original embedding matrix, LSAN leverages the notion of compositional embeddings, where each item embedding is composed by merging a group of selected base embedding vectors derived from substantially smaller embedding matrices. Meanwhile, to account for the intrinsic dynamics of each item, we further propose a temporal context-aware embedding composition scheme. Besides, we develop an innovative twin-attention network that alleviates the redundancy of the traditional multi-head self-attention while retaining full capacity for capturing long- and short-term (i.e., global and local) item dependencies. Comprehensive experiments demonstrate that LSAN significantly advances the accuracy and memory efficiency of existing sequential recommenders.

Deep learning and symbolic learning are two frequently employed methods in Sequential Recommendation (SR). Recent neural-symbolic SR models demonstrate their potential to enable SR to be equipped with concurrent perception and cognition capacities. However, neural-symbolic SR remains a challenging problem due to open issues like representing users and items in logical reasoning. In this paper, we combine the Deep Neural Network (DNN) SR models with logical reasoning and propose a general framework named Sequential Recommendation with Probabilistic Logical Reasoning (short for SR-PLR). This framework allows SR-PLR to benefit from both similarity matching and logical reasoning by disentangling feature embedding and logic embedding in the DNN and probabilistic logic network. To better capture the uncertainty and evolution of user tastes, SR-PLR embeds users and items with a probabilistic method and conducts probabilistic logical reasoning on users' interaction patterns. Then the feature and logic representations learned from the DNN and logic network are concatenated to make the prediction. Finally, experiments on various sequential recommendation models demonstrate the effectiveness of the SR-PLR. Our code is available at https://github.com/Huanhuaneryuan/SR-PLR.

Self-attention models have achieved the state-of-the-art performance in sequential recommender systems by capturing the sequential dependencies among user–item interactions. However, they rely on adding positional embeddings to the item sequence to retain the sequential information, which may break the semantics of item embeddings due to the heterogeneity between these two types of embeddings. In addition, most existing works assume that such dependencies exist solely in the item embeddings, but neglect their existence among the item features. In our previous study, we proposed a novel sequential recommendation model, i.e., MLP4Rec, based on the recent advances of MLP-Mixer architectures, which is naturally sensitive to the order of items in a sequence because matrix elements related to different positions of a sequence will be given different weights in training. We developed a tri-directional fusion scheme to coherently capture sequential, cross-channel, and cross-feature correlations with linear computational complexity as well as much fewer model parameters than existing self-attention methods. However, the cascading mixer structure, the large number of normalization layers between different mixer layers, and the noise generated by these operations limit the efficiency of information extraction and the effectiveness of MLP4Rec. In this extended version, we propose a novel framework – SMLP4Rec for sequential recommendation to address the aforementioned issues. The new framework changes the flawed cascading structure to a parallel mode, and integrates normalization layers to minimize their impact on the model’s efficiency while maximizing their effectiveness. As a result, the training speed and prediction accuracy of SMLP4Rec are vastly improved in comparison to MLP4Rec. Extensive experimental results demonstrate that the proposed method is significantly superior to the state-of-the-art approaches. The implementation code is available online to ease reproducibility.

Sequential recommendation systems which aim to predict the users’ future clicking items are essential in both research and industry services. Based on the contextual characteristics of users, a large number of existing methods exploit abundant information from user behavior sequences and gain users’ interests through expressive sequential models like recurrent neural networks and self-attention mechanism to predict the next item. However, the temporal generative process of behavior sequences, which is crucial to the complex relationships and distributions of users’ dynamic preferences, is rarely considered. In this paper, we propose HawRec (Hawkes process based sequential recommendation), which is a new representation learning approach to model the interacted sequences of users from a temporal point process perspective. In particular, we leverage temporal self-attention mechanism to make an analogy to the self-exciting mechanism of Hawkes process. We bring expressive superiority of neural network into the Hawkes process with non-parametric conditional intensity kernel, resulting in more accurate recommendation and inheriting robustness from the temporal point process. Furthermore, extensive experiments conducted on four public benchmarks demonstrate that our HawRec outperforms various baselines of state-of-the-art sequential recommendation.

Over the last few years, sequential recommender systems have achieved a great success in different applications. In the literature, it is generally believed that items farther away from the recommendation time have a weaker impact on the recommendation results. However, simply considering the distance between the interaction time and the recommendation time would prevent effective user representations. To solve this issue, we propose a Time Interval Aware Collaborative (TIAC) model with self-supervised learning for sequential recommendation. We propose to adjust the attention score learned from the time interval between an interaction time and the recommendation time using a time kernel function to achieve a better user representation. We also introduce self-supervised learning to combine the collaborative information obtained from a graph convolutional network and the sequential information learned from gated recurrent units to further enrich user representation. Extensive experiments on four real-world benchmark datasets show that our proposed TIAC model consistently outperforms state-of-the-art models under various evaluation metrics.KeywordsSequential recommendationAttention mechanismSelf-supervised learningGraph convolutional network

Recent advancements in generative models have positioned them as one of the principal tools for sequential recommendation due to their exceptional sample diversity and generalization capabilities. Among these, diffusion model-based sequential recommenders have achieved remarkable success. However, most existing approaches still face critical challenges, resulting in suboptimal generation quality: (1) They fail to leverage multimodal knowledge for constructing item representations with well-structured distributional characteristics and semantically enriched information; (2) They predominantly rely on discrete diffusion processes, leading to high error accumulation, reduced time efficiency, and constrained controllability in generative sampling. To mitigate these challenges, we propose LSGM4Rec, a novel framework that integrates Large Language Models (LLMs) with advanced multimodal encoding models to establish multimodal fusion embeddings for items. This design ensures distinct distributional characteristics while enabling the incorporation of semantically rich modal features into guidance condition. Furthermore, we pioneer the stochastic differential equations (SDEs) for recommendation, facilitating smooth transitions between data distributions and enabling optimal trade-off between sampling efficiency and generation quality. Extensive experiments on three datasets demonstrate that LSGM4Rec outperforms existing state-of-the-art sequential recommendation methods.

Sequential recommender systems (SRS) have gained increasing popularity due to their remarkable proficiency in capturing dynamic user preferences. In the current setup of SRS, a common configuration is to uniformly consider each historical behavior as a positive interaction. However, this setting has the potential to yield sub-optimal performance as each individual item often have a different impact on shaping the user's interests. Hence, in this paper, we propose a novel automatic sampling framework for sequential recommendation, named AutoSAM, to non-uniformly treat historical behaviors. Specifically, AutoSAM extends the conventional SRS framework by integrating an extra sampler to intelligently discern the skew distribution of the raw input, and then sample informative sub-sets to build more generalizable SRS. To tackle the challenges posed by non-differentiable sampling actions and to introduce multiple decision factors for sampling, we further design a novel reinforcement learning based method to guide the training of the sampler. Furthermore, we theoretically devise multi-objective sampling rewards including Future Prediction and Sequence Perplexity, and then optimize the whole framework in an end-to-end manner by combining the policy gradient. We conduct extensive experiments on benchmark recommendation models and four real-world datasets. The experimental results demonstrate the effectiveness of the proposed AutoSAM.

Sequential recommendation (SR) models based on Transformers have achieved remarkable successes. The self-attention mechanism of Transformers for computer vision and natural language processing suffers from the oversmoothing problem, i.e., hidden representations becoming similar to tokens. In the SR domain, we, for the first time, show that the same problem occurs. We present pioneering investigations that reveal the low-pass filtering nature of self-attention in the SR, which causes oversmoothing. To this end, we propose a novel method called Beyond Self-Attention for Sequential Recommendation (BSARec), which leverages the Fourier transform to i) inject an inductive bias by considering fine-grained sequential patterns and ii) integrate low and high-frequency information to mitigate oversmoothing. Our discovery shows significant advancements in the SR domain and is expected to bridge the gap for existing Transformer-based SR models. We test our proposed approach through extensive experiments on 6 benchmark datasets. The experimental results demonstrate that our model outperforms 7 baseline methods in terms of recommendation performance. Our code is available at https://github.com/yehjin-shin/BSARec.

Most sequential recommendation (SR) systems employing graph neural networks (GNNs) only model a user's interaction sequence as a flat graph without hierarchy, overlooking diverse factors in the user's preference. Moreover, the timespan between interacted items is not sufficiently utilized by previous models, restricting SR performance gains. To address these problems, we propose a novel SR system employing a hierarchical graph neural network (HGNN) to model factorial user preferences. Specifically, a timespan-aware sequence graph (TSG) for the target user is first constructed with the timespan among interacted items. Next, all original nodes in TSG are softly clustered into factor nodes, each of which represents a certain factor of the user's preference. At last, all factor nodes' representations are used together to predict SR results. Our extensive experiments upon two datasets justify that our HGNN-based factorial user modeling obtains better SR performance than the state-of-the-art SR models.

The convergence of machine learning (ML) and applied neuroscience continues to accelerate, driven by the synergistic demands of intelligent systems and deepening insights into the human nervous system. Building upon the success of Machine Learning and Applied Neuroscience: Volume I [https://www.frontiersin.org/journals/neurorobotics/articles/10.3389/ fnbot.2023.1191045/full], this second volume brings together cutting-edge research that exemplifies how computational intelligence-particularly deep learning, selfsupervision, and generative modeling-can address complex challenges in neurorobotics, neurorehabilitation, and behavior-aware intelligent systems.The four contributions in this Research Topic span a compelling spectrum: from the diagnosis of gait dysfunction in stroke survivors using cost-sensitive classifiers, to the generation of lifelike 3D human motion through generative adversarial networks (GANs), to next-generation sequential recommendation systems that model multigranularity behavior and feature interactions. Though seemingly diverse, these works share a unifying vision: leveraging advanced ML not only to model neural or behavioral data more accurately, but to extract clinically or functionally meaningful signals that empower real-world applications.One axis of innovation lies in clinical decision support through interpretable and robust ML. In their study, "Machine learning-based gait adaptation dysfunction identification using CMill-based gait data", Yang et al. (2024) (https://www.frontiersin.org/journals/neurorobotics/articles/10.3389/ fnbot.2024.1421401/full) tackle gait adaptation dysfunction (GAD)-a pervasive yet under-assessed impairment in post-stroke patients. Using data from an augmentedreality CMill treadmill, they extract kinematic and adaptability features across four ecologically valid tasks (e.g., obstacle avoidance, slalom walking). Among five classifiers evaluated, the AdaCost algorithm-designed to handle class imbalance and misclassification costs-achieved the best sensitivity (80%) and AUC (0.75). Crucially, feature importance analysis revealed that obstacle avoidance success and gait speed were the top predictors, aligning with clinical intuition and offering actionable biomarkers for rehabilitation planning. This work demonstrates how thoughtful integration of domain-aware data collection and cost-sensitive learning can yield deployable diagnostic aids.Parallel advances emerge in synthetic data generation for human-motion understanding. Wang et al. (2024), entitled "3D human pose data augmentation using Generative Adversarial Networks for robotic-assisted movement quality assessment" (https://www.frontiersin.org/journals/neurorobotics/articles/10.3389/ fnbot.2024.1371385/full) introduce a novel GANs-SVM-DenseNet pipeline to augment 3D human pose datasets-addressing a persistent bottleneck in training data scarcity and limited motion diversity. Their framework uses robotic-assisted capture for highfidelity grounding, GANs to generate realistic and varied motion sequences, DenseNet for hierarchical feature extraction, and SVM for precise motion-quality classification. Evaluated across four benchmarks (Human3.6M, MPI-INF-3DHP, NTU RGB+D, HumanEva), the model outperforms state-of-the-art methods in both accuracy (>96% on Human3.6M) and efficiency (30% faster inference). By closing the loop between data synthesis, feature learning, and quality assessment, this approach paves the way for scalable, robot-in-the-loop systems in sports science, rehabilitation, and virtual reality.Complementing these human-centered applications, two articles push the frontiers of sequential modeling in behavior-aware AI, with implications for neuroscience-inspired user modeling. Zhu et al. (2024a), at "Multi-granularity contrastive learning model for next POI recommendation" (https://www.frontiersin.org/journals/neurorobotics/articles/10.3389/ fnbot.2024.1428785/full), propose MGCL (Multi-Granularity Contrastive Learning) for next Point-of-Interest (POI) recommendation. Recognizing that user mobility is expressed not only at the location level but also through regions and categories, MGCL constructs multi-granular sequences and applies contrastive learning to encourage mutual enhancement across granularities. Experiments on three real-world datasets show consistent gains over 11 baselines-validating that modeling collaborative signals across abstraction levels mitigates data sparsity and enriches preference representation. This principle resonates with hierarchical processing in the brain, where sensory inputs are integrated across spatial and categorical scales.Extending this theme, the same team presents FIDS (Feature Interaction Dual Self-Attention Network) in the article entitled "Feature Interaction Dual Self-attention network for sequential recommendation" (https://www.frontiersin.org/journals/neurorobotics/articles/10.3389/ fnbot.2024.1456192/full), which challenges the assumption that item features are independent. FIDS first uses intra-item self-attention to model higher-order feature interactions (e.g., between brand, category, and seller in e-commerce), then employs dual self-attention streams to capture sequential dependencies both in item sequences and in the derived integrated-feature sequences. This architecture outperforms strong baselines-including FDSA and SASRec-by up to 9% in HR@5 on the Tmall dataset, proving that explicit modeling of feature interdependence enhances behavioral prediction. Such architectures may inform computational models of cognitive binding, where disparate perceptual attributes are fused into coherent percepts.Collectively, these works illustrate a maturing field where ML and neuroscience coevolve: ML architectures grow more neurobiologically plausible (e.g., through hierarchy, attention, and contrastive learning), while neuroscience and clinical applications benefit from increasingly nuanced, robust, and interpretable models.Looking ahead, key challenges remain-particularly around model explainability, realtime deployment in assistive robotics, cross-population generalizability, and ethical handling of behavioral data. Yet the trajectory is clear: the fusion of machine intelligence and applied neuroscience will continue to yield technologies that are not only more intelligent but also more human-centered.

IPSRM: An intent perceived sequential recommendation model