Feature-level Deeper Self-Attention Network for Sequential Recommendation

Zhang, Tingting; Zhao, Pengpeng; Liu, Yanchi; Sheng, Victor S.; Xu, Jiajie; Wang, Deqing; Liu, Guanfeng; Zhou, Xiaofang

doi:10.24963/ijcai.2019/600

Cited by 218 publications

(95 citation statements)

References 8 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Attribute-aware sequential models such as TransFM [16] and FDSA [29] leverage the contextual features to improve the sequential recommender models, in which these features are treated as auxiliary information to enhance the representation of items or sequences. In our S 3 -Rec, the L AAP loss and L M AP loss aim to fuse attribute with items or sequential contexts, which is able to achieve the same effect as previous methods [16,29]. Besides, the pre-trained data representations can be also applied to improve existing methods.…”

Section: Discussionmentioning

confidence: 99%

See 1 more Smart Citation

S3-Rec: Self-Supervised Learning for Sequential Recommendation with Mutual Information Maximization

Zhou

Wang

Zhao

et al. 2020

Proceedings of the 29th ACM International Conference on Information &Amp; Knowledge Management

542

321

View full text Add to dashboard Cite

Recently, significant progress has been made in sequential recommendation with deep learning. Existing neural sequential recommendation models usually rely on the item prediction loss to learn model parameters or data representations. However, the model trained with this loss is prone to suffer from data sparsity problem. Since it overemphasizes the final performance, the association or fusion between context data and sequence data has not been well captured and utilized for sequential recommendation. To tackle this problem, we propose the model S 3-Rec, which stands for Self-Supervised learning for Sequential Recommendation, based on the self-attentive neural architecture. The main idea of our approach is to utilize the intrinsic data correlation to derive self-supervision signals and enhance the data representations via pre-training methods for improving sequential recommendation. For our task, we devise four auxiliary self-supervised objectives to learn the correlations among attribute, item, subsequence, and sequence by utilizing the mutual information maximization (MIM) principle. MIM provides a unified way to characterize the correlation between different types of data, which is particularly suitable in our scenario. Extensive experiments conducted on six real-world datasets demonstrate the superiority of our proposed method over existing state-of-the-art methods, especially when only limited training data is available. Besides, we extend our self-supervised learning method to other recommendation models, which also improve their performance. CCS CONCEPTS • Information systems → Recommender systems. † Equal contribution.

show abstract

Section: Discussionmentioning

confidence: 99%

“…(10) SASRec F is our extension of SASRec, which concatenates the representations of item and attribute as the input to the model. (11) FDSA [29] constructs a feature sequence and uses a featurelevel self-attention block to model the feature transition patterns. This is the state-of-the-art model in sequential recommendation.…”

Section: Baseline Modelsmentioning

confidence: 99%

S3-Rec: Self-Supervised Learning for Sequential Recommendation with Mutual Information Maximization

Zhou

Wang

Zhao

et al. 2020

Proceedings of the 29th ACM International Conference on Information &Amp; Knowledge Management

542

321

View full text Add to dashboard Cite

show abstract

“…For instance, Shani et.al [17] examine the relation between a pair of items and then predict the probability of the next set of items for interaction. Attention mechanism is also applied in SRSs, FDSA for instance, use this technique to utilize transition patterns between features of items [18].…”

Section: A General Recommender Systemsmentioning

confidence: 99%

Towards a Deep Attention-Based Sequential Recommender System

et al. 2020

Self Cite

View full text Add to dashboard Cite

With the availability of a large amount of user-generated online data, discovering users' sequential behaviour has become an integral part of a Sequential Recommender System (SRS). Combining the recent observed items (i.e., short-term preferences) with prior interacted items (i.e., long-term preferences) has gained increasing attention in recent years. However, the existing methods mostly assume that all the adjacent items in a sequence are highly dependent, which may not be practical in real-world scenarios due to the uncertainty of customers' shopping behaviours. A user-item interaction sequence may contain some irrelevant items which may in turn lead to false dependencies between items. Moreover, current studies usually assign a static representation to each item when modeling a user's long-term preferences. Therefore, they cannot differentiate the contributions of the items. Specifically, these two types of users' preferences have been separately modeled and then linearly combined, which may fail to model complicated user-item interactions. In order to overcome the above mentioned problems, we propose a novel Deep Attention-based Sequential (DAS) model. DAS consists of three different blocks, (i) an embedding block: which embeds users and items into low-dimensional spaces; (ii) an attention block: which aims to discriminatively learn dependencies among items in both users' long-term and short-term item sets; and (iii) a fully-connected block: which first learns a mixture of users' preferences representation through a nonlinear way and then combines it with users' embeddings to have a personalized recommendation. Extensive experiments demonstrate the superiority of our proposed model compared to the state-of-the-art approaches in SRSs.

show abstract

“…Unlike the original LSTM model, we modified the LSTM by introducing the attentive embedding of item's image I uv into the generation of words. Suppose the word list of user u commenting on item i is w uv = {w 1 uv , w 2 uv , . .…”

Section: Review-driven Visual Attention Modelmentioning

confidence: 99%

“…Nowadays, when buying fashion products online, the user's decisions are primarily affected by the appearance of products [1]. However, the invisible features that the user cannot observe from the image, such as the material and quality of the clothes, also affect the user's decisions.…”

Section: Introductionmentioning

confidence: 99%

Visual and Textual Jointly Enhanced Interpretable Fashion Recommendation

Cui

2020

IEEE Access

Self Cite

View full text Add to dashboard Cite

With the rapid development of online shopping, interpretable personalized fashion recommendation using image has attracted increasing attention in recent years. The current work has been able to capture the user's preferences for visible features and provide visual explanations. However, they ignored the invisible features, such as the material and quality of the clothes, and failed to offer textual explanations. To this end, we propose a Visual and Textual Jointly Enhanced Interpretable (VTJEI) model for fashion recommendations based on the product image and historical review. The VTJEI can provide more accurate recommendations and visual and textual explanations through the joint enhancement of textual information and visual information. Specifically, we design a bidirectional two-layer adaptive attention review model to capture the user's visible and invisible preferences to the target product and provide textual explanations by highlighting some words. Moreover, we propose a review-driven visual attention model to get a more personalized image representation driven by the user's preference obtained from the historical review. In this way, we not only realize the joint enhancement of visual information and textual information but also provide a visual explanation by highlighting some regions. Finally, we performed extensive experiments on real datasets to confirm the superiority of our model on Top-N recommendations. We also built a labeled dataset for evaluating our provided visible and invisible explanations quantitatively. The result shows that we can not only provide more accurate recommendations but also can provide both visual and textual explanations. INDEX TERMS Explainable recommendation, fashion recommendation, visual and textual explanations.

show abstract

Feature-level Deeper Self-Attention Network for Sequential Recommendation

Cited by 218 publications

References 8 publications

S3-Rec: Self-Supervised Learning for Sequential Recommendation with Mutual Information Maximization

S3-Rec: Self-Supervised Learning for Sequential Recommendation with Mutual Information Maximization

Towards a Deep Attention-Based Sequential Recommender System

Visual and Textual Jointly Enhanced Interpretable Fashion Recommendation

Contact Info

Product

Resources

About