Green, Pervasive, and Cloud Computing: 15th International Conference, GPC 2020, Xi'an, China, November 13–15, 2020, Proceedings

Preface
Organization
Contents
Device-Free Sensing
Wavelet Analysis Based Noncontact Vital Signal Measurements Using mm-Wave Radar
1 Introduction
2 FMCW Radar Signal
3 Noncontact Vital Signals Detection Program
3.1 Theory of FMCW Radar Vital Signals Measurement
3.2 Instantaneous Vital Signals Detection Algorithm
4 Experiments and Results
4.1 Experimental Equipment and Parameters
4.2 Measurement Results
5 Conclusion
References
Study on Feasibility of Remote Metal Detection Using Millimeter Wave Radar for Convenient and Efficient Security Check
1 Introduction
2 Related Work
3 Overview
4 Methodology
4.1 Preliminaries: Radar Fundamental
4.2 The Establishment of Human Body Model
4.3 Metal Detection
5 Experiment
5.1 Dataset
5.2 Implementation Details
5.3 System Performance
5.4 System Practicability
6 Discussion and Future Work
7 Conclusion
References
Long-Range Gesture Recognition Using Millimeter Wave Radar
1 Introduction
2 Related Work
3 Overview
4 Methodology
4.1 Point Cloud Model of Gesture
4.2 The Model of Convolutional Neural Network
5 Experiment
5.1 Implementation Details
5.2 Performance Analysis
5.3 Forecast Accuracy for New Users
5.4 Practical Application Analysis
6 Discussion
6.1 Distance Influence
6.2 Multiple People Influence
6.3 Real-Time Consideration
7 Conclusion
References
Towards Fine-Grained Indoor White Space Sensing
1 Introduction
2 Indoor TV Spectra Measurement
2.1 Equipment and Setup
2.2 Synchronous Measurement
2.3 Asynchronous Measurement
3 System Design
3.1 System Overview
3.2 Multitask Gaussian Process Based Spatial Interpolation
3.3 Determining the Positions of Candidate Locations
4 Performance Evaluation
4.1 Evaluation Setup
4.2 Performance of Gaussian Process Based Interpolation
4.3 Different Positions of Candidate Locations
5 Related Work
6 Conclusion
References
Evaluating mmWave Sensing Ability of Recognizing Multi-people Under Practical Scenarios
1 Introduction
2 Related Works
2.1 Application Prospect of MmWave
2.2 Recognition Using mmWave Signals
3 System Design and Experimental Process
3.1 Overview of Main Experiment
3.2 Hardware Configuration and Data Acquisition
4 Experiment Evaluation and Analysis
4.1 Basic Working Performance of mmWave FMCW Radar
4.2 Minimum Effective Clustering Distance Between Two Persons
4.3 Occlusion Relationship Effect
5 Limitations
5.1 Reflecting by Flat Surfaces
5.2 Scope of Monitoring
6 Conclusion
References
Machine Learning
GradSA: Gradient Sparsification and Accumulation for Communication-Efficient Distributed Deep Learning
1 Introduction
2 Background and Problem Setting
3 Design Overview
4 Layer-Level Gradient Sparsification
5 Staleness-Compensated Historical Gradient Accumulation
5.1 Historical Gradient Approximation
5.2 Historical Gradient Accumulation
6 Evaluation
6.1 Experiment Setup
6.2 Overall Training Performance Comparison
6.3 Scalability Evaluation
7 Related Work
8 Conclusion
References
An Improved Artificial Bee Colony Algorithm with Multiple Search Strategy
1 Introduction
2 Original Artificial Bee Colony Algorithm
3 The Proposed ABC Algorithm
3.1 Initialization Phase
3.2 Employed Bees Phase
3.3 Onlooker Bees Phase
3.4 Scout Bee Phase
4 Validation and Comparison
4.1 Benchmark Functions
4.2 Simulation and Results
5 Conclusion
References
An Efficient Data Prefetch Strategy for Deep Learning Based on Non-volatile Memory
1 Introduction
2 Related Work
2.1 Optimizations for Data Bottleneck in DL Systems
2.2 Development of NVRAM
3 NVRAM-Based Cache Sliding Strategy for Data Prefetch
3.1 NVRAM Data Initialization
3.2 NVRAM Data Sliding
3.3 NVRAM Refilling for Next Sliding Stage
4 Evaluation
4.1 Experiment Environment
4.2 Performance Evaluation
5 Conclutions and Future Work
References
A Drift Detection Method Based on Diversity Measure and McDiarmid’s Inequality in Data Streams
1 Introduction
2 The Diversity Measure and McDiarmid Drift Detection Method
2.1 Diversity Measure
2.2 Drift Detection Method
2.3 McDiarmid’s Inequality
2.4 Parameters Analysis
3 Experiment Evaluation
4 Conclusion
References
An Improved Sparse Representation Classifier Based on Data Augmentation for Time Series Classification
1 Introduction
2 SRC and ESRC
3 Improved Extend SRC
4 Experimental Results
4.1 Experiments on UCR Archive
4.2 Experiments on Predicting the Effect of Aerobic Exercise Intervention on 24 Young Patients with Hypertension
5 Conclusion
References
Concept Stability Based Isolated Maximal Cliques Detection in Dynamic Social Networks
1 Introduction
2 Preliminaries
2.1 Clique
2.2 FCA
2.3 Concept Stability
3 Problem Statement
4 Concept Stability Based Isolated Maximal Clique Detection
4.1 Formal Context Construction from Dynamic Social Network
4.2 Isolated Maximal Clique Detection
5 Experiments
5.1 Data Set and Configurations
5.2 Experimental Results
6 Conclusions
References
Echo State Network Based on L0 Norm Regularization for Chaotic Time Series Prediction
1 Introduction
2 Preliminaries
3 ESN Based on L0 Norm Regularization
3.1 Orthogonal Matching Pursuit (OMP)
3.2 OMP-ESN Algorithm
4 Simulations
4.1 Datasets Description
4.2 Experimental Setup
4.3 Results Discussion
5 Conclusion
References
Recommendation Systems
MI-KGNN: Exploring Multi-dimension Interactions for Recommendation Based on Knowledge Graph Neural Networks
1 Introduction
2 Related Work
3 MI-KGNN
3.1 Problem Definition
3.2 Problem Analysis
3.3 Implementation of MI-KGNN
4 Experiments
4.1 Dataset Description
4.2 Baselines
4.3 Experimental Settings
4.4 Results and Discussion
5 Conclusions and Future Work
References
MGCN4REC: Multi-graph Convolutional Network for Next Basket Recommendation with Instant Interest
1 Introduction
2 Related Work
2.1 Representation Learning
2.2 Sequential Recommendation System
3 MGCN4REC Framework
3.1 Preliminaries
3.2 Multi-graph-Based User-Item Representation
3.3 User Preferences and Interests Modeling
3.4 Attention-Based Preferences Aggregating
3.5 Model Training
4 Experiment
4.1 Dataset
4.2 Baselines
4.3 Evaluation Metric
4.4 Experimental Results
5 Conclucion
References
Urban Computing
Using Deep Active Learning to Save Sensing Cost When Estimating Overall Air Quality
1 Introduction
2 Related Work
3 Problem Description
4 Data Preprocessing
4.1 Data Description and Missing Data Processing
4.2 Data Feature Processing
5 Proposed Method
5.1 Acquisition Function for the First Phase of Active Learning
5.2 The Second Phase of Active Learning
5.3 Active Variational Adversarial Model
6 Experiments
6.1 Learning of AQI Level
6.2 Active Learning of AQI Level
7 Conclusion
References
Interpretable Multivariate Time Series Classification Based on Prototype Learning
1 Introduction
2 Related Work
2.1 Multivariate Time Series Classification
2.2 Interpretability
3 Method
3.1 Model Architecture Overview
3.2 Data Encoding
3.3 Prototype Learning
3.4 Fully Connection
4 Experiment
4.1 Experimental Settings
4.2 Result
4.3 Interpretation
5 Conclusion
References
A Driver-Centric Vehicle Reposition Framework via Multi-agent Reinforcement Learning
1 Introduction
2 Related Works
3 Problem Statement
4 Methodologies
4.1 Multi-agent DDPG
4.2 Geographic-Based MADDPG
5 Experiments
5.1 Simulator
5.2 Performance Evaluation
5.3 Consideration of ORR
5.4 Passenger Satisfaction
5.5 Ablation Study
6 Conclusions
References
Demand-Responsive Windows Scheduling in Tertiary Hospital Leveraging Spatiotemporal Neural Networks
1 Introduction
2 Preliminary and Framework Overview
2.1 Preliminaries
2.2 Framework Overview
3 Hospital Visit Demand Graph Modeling
3.1 Taxi Trajectory and Hospital Location Based Hospital Visit Demand Extraction
3.2 Spatiotemporal Hospital Visit Demand Graph Modeling
4 Registration Windows Scheduling
4.1 ST-GNN Based Hospital Visit Demand Forecasting
4.2 Demand-Responsive Registration Windows Scheduling
5 Evaluation
5.1 Dataset Description
5.2 Evaluation on Visit Forecast
5.3 Case Studies on Windows Scheduling
6 Relative Work
6.1 Spatiotemporal Sequence Forecast
6.2 Graph Neural Networks
7 Conclusion and Future Work
References
Prediction Technology for Parking Occupancy Based on Multi-dimensional Spatial-Temporal Causality and ANN Algorithm
1 Introduction
2 Related Work
2.1 Real-Time On-Line Prediction
3 Prediction Methodology
3.1 Multidimensional Spatial-Temporal Causality Model
3.2 Parking Occupancy Prediction
4 Analysis of Experimental Results
4.1 Data Selection and Normalization
4.2 Neural Network Construction
4.3 Analysis of Experimental Results
5 Conclusion
References
An Improved Leaky-ESN for Electricity Load Forecasting
1 Introduction
2 Related Work
3 Preliminary
4 AESN for Electricity Load Forecasting
4.1 The Modular Control Strategy
4.2 AESN-I with Initial Sparsely-Connected Reservoir
4.3 AESN-II with Initial Fully-Connected Reservoir
5 Experiments and Results
5.1 Dataset Description
5.2 Experimental Settings
5.3 Results Discussion
6 Conclusions and Future Work
References
Human Computer Interaction
MateBot: The Design of a Human-Like, Context-Sensitive Virtual Bot for Harmonious Human-Computer Interaction
1 Introduction
2 Related Work
3 The MateBot Architecture
4 MateBot: Detailed Design
4.1 Context-Sensitive Network
4.2 Appearance Translation Network
4.3 Dialogue Network
5 Evaluation
5.1 Dataset
5.2 Evaluation Results
6 Conclusion
References
User Behavior Analysis Toward Adaptive Guidance for Machine Operation Tasks
1 Introduction
2 Related Work
3 Key Idea
4 Operational Behavior Detection
4.1 Visual Features
4.2 Behavioral Features
4.3 Correlation to Skill Level and Difficulty
5 Experimental Results
5.1 Experimental Conditions
5.2 Behavior Changes Through Skill Improvements
6 Discussion
7 Conclusion
References
Behavior Fingerprints Based Smartphone User Authentication: A Review
1 Introduction
2 Behavioral Biometrics Authentication Solutions on Smartphones
2.1 Uni-Modal Behavioral Authentication Systems
2.2 Multimodal Behavioral Authentication Systems
3 Discussion
4 Open Issues and Challenges
5 Conclusion
References
A System to Find the Change of One's Vision Implicitly
1 Introduction
2 Motivation and Challenges
2.1 Motivation
2.2 Challenges and Approach
3 System Overview
4 Activity Recognition
5 Eye-Screen Distance Estimation
5.1 Eye Localization
5.2 Eye-Screen Distance
5.3 Find Outliers
5.4 Median as Representation
5.5 Least Squares
6 Implementation and Evaluation
6.1 Dataset
6.2 Detection Result
7 Discussion
8 Conclusions
References
Internet of Things and Edge Computing
Energy-Aware Marginal Multi-attribute Federated Query in IoT Networks
1 Introduction
2 Energy Model
3 Marginal Multi-attribute Federated Query
3.1 Cost Calculation of Queried Packets Transmitting for Marginal Edge Nodes
3.2 Marginal Multi-attribute Federated Query Model
4 Implementation and Evaluation
4.1 Baseline Methods
4.2 Experiment Settings
4.3 Evaluation Results
5 Related Works and Comparison
5.1 Energy-Aware Routing Query
5.2 Spatial Keyword Query
6 Conclusion
References
A Novel UAV Charging Scheme for Minimizing Coverage Breach in Rechargeable Sensor Networks
1 Introduction
2 Problem Description
2.1 Network Model and Assumptions
2.2 Partial Recharging Model
2.3 Problem Definition
3 Complete Coverage and Energy Knowledge Partial Charging Scheme (Co-EPaCS)
3.1 Finding a Subset of Sensors to Charge and Deciding When to Start a Charging Tour
3.2 Planning UAV Charging Tour and Charging the Sensors
4 Performance Evaluation
4.1 Parameter Settings
4.2 Results and Discussions
5 Conclusion
References
A Parallel Tasks Scheduling Algorithm with Markov Decision Process in Edge Computing
1 Introduction
2 Related Work
2.1 Research on Edge Computing
2.2 Applications of Reinforcement Learning in Scheduling Problems
3 Problem Definition
4 Our Method
4.1 Algorithm Overview
4.2 Establishment of Batch
4.3 Q-learning Method
5 Experiment
5.1 Experiment Settings
5.2 Effectiveness Evaluation
6 Conclusion
References
Positioning
Localization Research Based on Low Cost Sensor
1 Introduction
2 System Overview
3 Agorithm Design
3.1 Flow Chart
3.2 Walking State Detection
3.3 Extended Kalman Filter Algorithm
4 Experimental Results and Analysis
5 Conclusion
References
Relative Floor Estimation for Indoor Co-navigation: A Machine Learning Approach
1 Introduction
2 Problem Description
3 Relative Floor Classification
3.1 Relative Feature
3.2 Relation Classifier
4 Experiments
4.1 Experiment Settings
4.2 Comparison of Different Classifiers
4.3 The Impact of Horizontal Distance by GPS
4.4 The Impact of Training Samples
4.5 The Impact of Different Signal Sources
5 Conclusion
References
Applications of Computer Vision
Defect Detection of Production Surface Based on CNN
1 Introduction
2 Related Work
3 Implement Details
3.1 Dataset
3.2 Data Augmentation
3.3 Backbone
3.4 Detector
3.5 Extra Module
3.6 Transfer Learning
4 Experiment
4.1 Experiment Setup
4.2 Multi-scale Anchors
4.3 Comparison
4.4 Analysis
5 Conclusion
References
An Image-Based Method for 3D Human Shapes Retrieval
1 Introduction
2 Method
2.1 The Spatial Coordinates of Human in the Image
2.2 3D Pose Estimation from 2D Human Pose
2.3 Similarity Computation for 3D Model Retrieval
3 Experiments
4 Conclusion
References
MobiVision: A Novel Energy-Efficient Mobile Deep Learning Framework for Computer Vision
1 Introduction
2 Related Work
3 Method
3.1 Solution Space
3.2 Large-Scale Solution Space Partitioning
3.3 Customized Neural Network Framework Designing
4 Model Performance
4.1 DataSets and Data Preprocessing
4.2 Coarse-Grained Clustering
4.3 Classifiers
5 Conclusion
References
CrowdSensing
Mobile Crowd-Sensing System Based on Participant Selection
1 Introduction
2 Relative Work
3 Problem Formulation
3.1 Definition
3.2 Assumptions
3.3 Problem Formulation
4 Method
4.1 Data Speculation
4.2 Participant Selection
5 Experiment
5.1 Data Preprocessing
5.2 State Setting Experiment
5.3 Air Quality Data Experiment
6 Conclusion
References
Quality-Aware and Penalty-Sensitive Opportunistic Crowdsensing in Mobile Relay Networks
1 Introduction
2 Related Works
3 Problem Formulation
3.1 System Overview
3.2 Relay Encouraging Coverage Maximization (RECM)
4 Mechanism Design
4.1 Basic-REG
4.2 Approximation Analysis
4.3 Enhanced-REG
4.4 Penalty Sensitive Scenario
5 Performance Evaluation
5.1 Simulation Setup
5.2 Evaluation of RECM
5.3 Evaluation of Penalty Sensitive Scenario
6 Conclusion
References
Cloud and Related Technologies
Failure Prediction with Hierarchical Approach in Private Cloud
1 Introduction
2 Related Work
3 System Architecture
3.1 Data Collection Module
3.2 Data Storage Module
3.3 VM Fault Monitoring Module
3.4 Processing Behavior Management
4 Proposed Method
4.1 Data Cleaning
4.2 Fast Filtering
4.3 Confidence Judgment
4.4 Complex Model
4.5 The HFP Algorithm
5 Result Analysis
5.1 Experiments Setup
5.2 Performance Analysis
5.3 Feature Analysis
6 Conclusions
References
Research on Job Scheduling Algorithms Based on Cloud Computing
1 Introduction
2 Related Work
3 Evaluation Model
3.1 Problem Definition
3.2 Deadline Estimation Model
4 (MQWAGS) Job Scheduling Algorithm
4.1 Deadline Estimation Model Design Goals
4.2 Examples of Job Scheduling Strategies
4.3 Multi-queue Load-Aware Greedy Scheduling (MQWAGS)
5 Experiment
5.1 Experimental and Simulation Setup
5.2 Experimental Results and Analysis and Simulation Setup
6 Conclusions
References
Recovering Cloud Services Using Hybrid Clouds Under Power Outage
1 Introduction
2 Model and Problem Formulation
3 Online Algorithm
3.1 Problem Transformation Using Lyapunov Optimization
3.2 Power-Aware Online Algorithm
4 Performance Evaluation
5 Conclusion
References
BED: A Block-Level Deduplication-Based Container Deployment Framework
1 Introduction
2 Background and Motivation
2.1 Deploying a Non-local Container
2.2 Large Image Size Caused by Data Redundancy
2.3 Deduplication for Container Image
3 System Design
3.1 Overview
3.2 Enabling Block-Level Deduplication During Pulling an Image
3.3 Reconstructing Image Layer with Blocks
3.4 Overlapping Decompression, Pull, and Reconstruction
4 Evaluation
4.1 Experimental Setups
4.2 Overall Performance When Pulling Images in Batch
4.3 Pull Time of Each Type of Images
4.4 Benefit of Overlapping
5 Related Work
5.1 Fast Deployment of Container
5.2 Deduplication in Virtualized Environments
6 Conclusion
References
Author Index