Deep Tensor Factorization for Multi-Criteria Recommender Systems

Chen, Zhengyu; Gai, Sibo; Wang, Donglin

doi:10.1109/bigdata47090.2019.9005677

Cited by 20 publications

(9 citation statements)

References 25 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Cross-domain meta-learner θ: MSN takes modulation vectors as input to learn a cross-domain multi-initialization metalearner for new multi-modal tasks. As lower layers in deep networks tend to produce general features while higher layers tend to produce specific features [11,12], the lower layers of MSN, named as common network θ c , are to learn domaininvariant knowledge while the higher layers of MSN, named as private networks θ d p for d ∈ {s, t} in Fig. 2, are to learn domain-specific knowledge.…”

Section: Meta Separation Network (Msn)mentioning

confidence: 99%

Multi-Initialization Meta-Learning with Domain Adaptation

Chen

Wang

2021

ICASSP 2021 - 2021 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)

Self Cite

View full text Add to dashboard Cite

Recently, meta learning providing multiple initializations has drawn much attention due to its capability of handling multi-modal tasks drawn from diverse distributions. However, because of the difference of class distribution between meta-training and meta-test domain, the domain shift occurs in multi-modal meta-learning setting. To improve the performance on multi-modal tasks, we propose multi-initialization meta-learning with domain adaptation (MIML-DA) to tackle such domain shift. MIML-DA consists of a modulation network and a novel meta separation network (MSN), where the modulation network is to encode tasks into common and private modulation vectors, and then MSN uses these vectors separately to update the cross-domain meta-learner via a double-gradient descent process. In addition, the regularization using inequality measure is considered to improve the generalization ability of the meta-learner. Extensive experiments demonstrate the effectiveness of our MIML-DA method to new multi-modal tasks.

show abstract

Section: Meta Separation Network (Msn)mentioning

confidence: 99%

Multi-Initialization Meta-Learning with Domain Adaptation

Chen

Wang

2021

ICASSP 2021 - 2021 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)

Self Cite

View full text Add to dashboard Cite

show abstract

“…Both users' content information and ontological semantic filtering are utilized to solve sparsity problem in [35]. Content features extracted by stacked denoising autoencoders are integrated into tensor factorization in [36]. Even though the work presented in [36] utilizes denoising autoencoders to extract features from content information, the work is still based on linear assumption during prediction process.…”

Section: Related Workmentioning

confidence: 99%

“…Content features extracted by stacked denoising autoencoders are integrated into tensor factorization in [36]. Even though the work presented in [36] utilizes denoising autoencoders to extract features from content information, the work is still based on linear assumption during prediction process. [37] propose to use preference-based similarity instead of computing correlations over sparse rating profiles to handle with sparsity issue.…”

Section: Related Workmentioning

confidence: 99%

A new similarity-based multi-criteria recommendation algorithm based on autoencoders

2021

Turk J Elec Eng & Comp Sci

View full text Add to dashboard Cite

Recommender systems provide their users an efficient way to handle with information overload problem by offering personalized suggestions. Traditional recommender systems are based on two-dimensional user-item preference matrix which constructed depending on the users' overall evaluations over items. However, they have begun to present their preferences over under various circumstances. Thus, traditional recommendation techniques fail to process multicriteria ratings during the recommendation process. Multi-criteria recommender systems are an extension of traditional recommender systems that utilize multi-criteria-based user preferences. Multi-criteria recommender systems provide more personalized and accurate predictions compared to traditional recommender systems. However, the increased amount of data dimension causes sparsity to be a major problem of such systems. Especially, the similarity-based multi-criteria recommender systems may fail to find similar neighbors to an active user due to the lack of co-rated items among users.Therefore, we propose a new similarity-based multi-criteria collaborative filtering approach based on autoencoders. In order to handle with sparsity, the proposed method extracts non-linear, low-dimensional, dense features from raw and sparse users'/items' preferences. Our experimental outcomes show that the proposed work can amortize the negative impacts of sparsity over the accuracy comparing with the state-of-the-art multi-criteria recommendation techniques.

show abstract

“…The DTF can extract hierarchical and meaningful features of multi-channel images such as hyperspectral images, thus is popular in image classification and pattern classification [227,228,229]. DTF can also be used in recommender systems [230], scene decomposition [231], and fault diagnosis [232].…”

Section: Dealing With High Dimensional Datamentioning

confidence: 99%

Deep Learning Methods for Solving Linear Inverse Problems: Research Directions and Paradigms

Bai,

Chen,

Chen

et al. 2020

Preprint

View full text Add to dashboard Cite

The linear inverse problem is fundamental to the development of various scientific areas. Innumerable attempts have been carried out to solve different variants of the linear inverse problem in different applications. Nowadays, the rapid development of deep learning provides a fresh perspective for solving the linear inverse problem, which has various well-designed network architectures results in state-of-the-art performance in many applications. In this paper, we present a comprehensive survey of the recent progress in the development of deep learning for solving various linear inverse problems. We review how deep learning methods are used in solving different linear inverse problems, and explore the structured neural network architectures that incorporate knowledge used in traditional methods. Furthermore, we identify open challenges and potential future directions along this research line.

show abstract

Deep Tensor Factorization for Multi-Criteria Recommender Systems

Cited by 20 publications

References 25 publications

Multi-Initialization Meta-Learning with Domain Adaptation

Multi-Initialization Meta-Learning with Domain Adaptation

A new similarity-based multi-criteria recommendation algorithm based on autoencoders

Deep Learning Methods for Solving Linear Inverse Problems: Research Directions and Paradigms

Contact Info

Product

Resources

About