A Generalized Nesterov's Accelerated Gradient-Incorporated Non-Negative Latent-Factorization-of-Tensors Model for Efficient Representation to Dynamic QoS Data

Dynamic Quality-of-Service (QoS) data can be efficiently represented by a Non-negative Latent-factorization-of-tensors model, which relies on a Non-negative and Multiplicative Update on Incomplete Tensors (NMU-IT) algorithm. Nevertheless, NMU-IT frequently encounters slow convergence and inefficient hyper-parameters selection. Targeting at overcome these critical defects, this paper proposed to improve the NMU-IT algorithm from two perspectives: a) integrating a generalized Nesterov's accelerated gradient method to accelerate the resultant model's convergence rate, and b) establishing the hyper-parameter adaptation mechanism through the particle swarm optimization strategy. On the basis of these conceptions, this study successfully builds a Generalized Nesterov's Accelerated Gradient-incorporated Non-negative Latent-factorization-of-tensors (GNL) model for precisely and high-efficiently representing the dynamic QoS data. The proposed GNL model has shown its superiority over several advanced models concerning both the precision of estimating missing QoS data and training efficiency, as demonstrated by the experiments conducted on two dynamic QoS datasets.

I. Introduction

Industrial applications related to cloud services commonly evaluate the performance of hired services through user-side non-functional Quality-of-Service (QoS) data, such as invoking expenses, throughput and response-time [1], [2]. Commonly, user-side QoS data can provide specifically personalized information as it varies over time. Hence, how to execute precise representation learning to dynamic QoS data has become a major issue during the last decade [3], [4], [5]. A high-dimensional and incomplete (HDI) tensor can fully model the temporal structure of dynamic QoS data, while it may fail to establish full connections among all involved nodes (i.e., users, services and concerned time slots), making an HDI tensor extremely incomplete [6], [7], [8], [9], [10], thus leading to great challenge to its precise representation learning [56].