Industrial Internet of Things (IIoT) is producing massive data which are valuable for knowing running status the underlying equipment. However, these involve various operation events that span some time, raise questions on how to model long memory states, and predict based historical accurately. This paper aims develop a method of: (1) analyzing equipment working condition sensed data; (2) building prediction forecasting designing deep neural network (3) improving accuracy by systematic...

With the surge of academic papers, it has become a common practice to recommend papers based on authors’ research interests. Existing methods focus leveraging author-paper interactions mine interests with coauthorship networks. However, sparse would pose huge challenge distinguish authors. Fortunately, inter-dependent knowledge across provides rich potential heterogeneous connections for interactions, offering much insights learning Therefore, we propose meta-relation guided coupling...

Video data has become the largest source of big data. Owing to video data's complexities, velocity, and volume, public security other surveillance applications require efficient, intelligent runtime processing. To address these challenges, a proposed framework combines two cloud-computing technologies: Storm stream processing Hadoop batch It uses deep learning realize intelligence that can help reveal knowledge hidden in An implementation this five architecture styles: service-oriented...

Although Big Data Cloud (e.g., MapReduce, Hadoop and Dryad) makes it easy to develop run highly scalable applications, efficient provisioning fine-tuning of these massively distributed systems remain a major challenge. In this paper, we describe general approach help address challenge, based on instrumentations dataflow-driven performance analysis. Based approach, have implemented HiTune, scalable, lightweight extensible analyzer for Hadoop. We report our experience how HiTune helps users...

Microservices are bundled and generating traffic on the backend systems that need to scale demand. When microservices generate variant unexpected, challenge is classify workload adjust scaling policy reflect resource demand timely accurately. In this paper, we propose a microservice architecture encapsulates functions of monitoring metrics learning pattern. Then service used predict future for decision making provisioning. We deploy two machine algorithms emulated by Netflix benchmark...

Traditional neural networks usually concentrate on temporal data in system simulation, and lack of capabilities to reason inner logic relations between different dimensions collected from embedded sensors. This paper proposes a graph network-based modeling approach for IoT equipment (called GNNM-IoT), which considers both data, vertices denote sensor edges relationships vertices. The GNNM-IoT model's sensors with produce nonlinear complex relationships. We have evaluated the using...

In this article, we follow a process of explainable artificial intelligence (XAI) method development and define two metrics in terms consistency efficiency guiding the evaluation XAI explanations.

Cloud monitoring is of a source big data that are constantly produced from traces infrastructures, platforms, and applications. Analysis delivers insights the system's workload usage pattern ensures workloads operating at optimum levels. The analysis process involves query extraction, analysis, result visualization. Since volume big, these operations require scalable reliable architecture to extract, aggregate, analyze in an arbitrary range granularity. Ultimately, results become knowledge...

Machine Learning Operations (MLOps) aim to establish a set of practices that put tools, pipelines, and processes build fast time-to-value machine learning development projects. The lifecycle project encompasses roles, stacks software frameworks multiple types computing resources. Such complexity makes MLOps support usually bundled with commercial cloud platforms is referred as vendor lock. In this paper, we provide an alternative solution devises platform open source on any virtual Our...

In this study, we propose the early adoption of Explainable AI (XAI) with a focus on three properties: Quality explanation, explanation summaries should be consistent across multiple XAI methods; Architectural Compatibility, for effective integration in XAI, architecture styles both methods and models to explained must compatible framework; Configurable operations, explanations are operable, akin machine learning operations. Thus, an reproducible tractable trustworthy. We present XAIport,...

This paper presents an enhanced efficiency 3-D convolution operator based on optimal field programmable gate array (FPGA) accelerator platform. The proposed system takes advantages of the intermediate data delay lines, implemented in FPGA, to avoid loading repetition input feature maps. performs 268.07 giga operations per second at 100-MHz operation frequency, with 330-mW power consumption. We experimentally demonstrate accelerator, comparison conventional technologies. may find interesting...

The Infrastructure as a Service (IaaS) cloud industry that relies on leasing virtual machines (VMs) has significant portion of business values finding the dynamic equilibrium between two conflicting phenomena: underutilization and surging congestion. Spot instance been proposed an elegant solution to overcome these challenges, with ultimate goal achieve greater profits. However, previous studies recent spot pricing schemes reveal artificial policies do not comply nature phenomena. Motivated...

Accurate fault localization renders software test resource allocation and maintenance cost-efficient. However, this is challenging when there are false alarm repercussions caused by module coupling of complex software. In article, therefore, we propose a new method for multiple from the perspective network spectrum based on graph neural model. First, constructed model under to represent relationships among modules theory. addition, suits were executed recorded construct program spectrum....

Software vulnerability detection aims to proactively reduce the risk software security and reliability. Despite advancements in deep-learning-based detection, a semantic gap still remains between learned features human-understandable semantics. In this paper, we present an XAI-based framework assess program code graph context as feature representations their effect on classification into multiple Common Weakness Enumeration (CWE) types. Our XAI is deep-learning-model-agnostic...

Self-supervised learning continues to drive advancements in machine learning. However, the absence of unified computational processes for benchmarking and evaluation remains a challenge. This study conducts comprehensive analysis state-of-the-art self-supervised algorithms, emphasizing their underlying mechanisms intricacies. Building upon this analysis, we introduce model-agnostic computation (UMAC) process, tailored complement modern algorithms. UMAC serves as global explainable artificial...

The characteristics of neo4j graph database has been introduced, several methods constructing based on importing data have compared and analyzed, the "neo4j-admin" method finally selected to import air crash data, realize visual display query. Specific implementation steps as follows:(1) Data sets worldwide since 1908 [1] cleaned preprocessed remove incomplete lines unimportant fields;(2) Crashed aircraft information table, accident details table above two association relationship designed...

Pervasive computing systems are heterogenous and complex as they usually involve human activities, various sensors actuators well middleware for system controlling. Therefore, analyzing such is highly non-trivial. In this work, we propose to use formal methods pervasive systems. Firstly, a modeling framework proposed cover main characteristics of (e.g., context-awareness, concurrent communications, layered architectures). Secondly, identify the safety requirements free deadlock conflicts...

Abstract In this article, an uncertain disturbance rejection control problem for the affine system in presence of asymmetric input constraints is addressed using event‐triggered method. The converted to optimal problem, and a zero‐sum game‐based method proposed solve problem. To deal with constraints, new cost function proposed. controller updated only when triggering condition satisfied, which can reduce computational complexity.In order obtain that minimizes performance index worst‐case...

In this paper we introduced a new R-tree node splitting algorithm. As an indexing technique for multi-dimensional data, is widely used in geographical information systems, CAD systems and spatial databases. An consists of nodes which turn consist records. Each must contain limited number records order that it can be stored within one disk block, thus algorithm while inserting record into full node. A the crucial factors query performance since bad splits would result inefficient structure....