Abstract
In recommendation systems (RSs), nowadays, not only the traditional user-item rating matrices but also more additional information like contents, contexts, trust friends and other auxiliary information are available to enhance the performance of RS, leading to content-aware, context-aware, trust-aware RS, etc. Thus, it provides much potential to take into consideration the additional information in RS. Hence, we focus on a general low-rank matrix factorization (LRMF) model with similarity constraints and propose a decentralized algorithm based on alternating direction method of multipliers (ADMM) to relieve the computation burden in each server while preserving privacy. What’s more, we utilize low-complexity skills in numerical analysis to reduce the computational complexity, based on the exploitation of the special form of the problem. Finally, simulations are performed to validate the effectiveness of our algorithms.
Highlights
Recommendation system (RS) aims to recommend the most suitable items to particular users, thereby reducing the information load and providing personalized services [1]
Relevant research can be traced back to mid-1990s when recommendations were made explicitly relying on the rating structure, and recommendation techniques focused on collaborative filtering (CF), content-based methods (CB) and hybrid methods [2]
The guidance obtained from the above analysis of different factors can be summarized as follows: i) our alternating direction method of multipliers (ADMM) algorithm shows good convergence; ii) there exists a tradeoff for top-k and number of features when taking the computational complexity and accuracy into consideration; iii) the sparsity of the data set affects the choice of training rate
Summary
Recommendation system (RS) aims to recommend the most suitable items to particular users, thereby reducing the information load and providing personalized services [1]. For decentralized algorithms, [27] gives an overview and offers algorithmic analysis at a high level of abstraction Among these decentralized algorithms, alternating direction method of multipliers (ADMM) is well studied and widely applied in diverse application domains including optimization and machine learning [28]–[31], sparsity and low-rank recovery [32], resource management [33], multicell coordinated beamforming [34] and average consensus problem [35]. Decentralized algorithm for RS with similarity constraints: A decentralized algorithm based on ADMM is designed for LRMF, which can relieve the computation burden in each agent and alleviate the robustness concerns raised by the centralized fashion while keeping data privacy. The n × n identity matrix is presented by In. ρ (·) represents the spectral norm. λ(A) and σ (A) denote the eigenvalues and singular values of A respectively
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
