Computational Intelligence and

Intelligent Systems 9th International

Symposium ISICA 2017 Guangzhou
China November 18 19 2017 Revised
Selected Papers Part II Kangshun Li
Visit to download the full and correct content document:
https://textbookfull.com/product/computational-intelligence-and-intelligent-systems-9th
-international-symposium-isica-2017-guangzhou-china-november-18-19-2017-revised
-selected-papers-part-ii-kangshun-li/
More products digital (pdf, epub, mobi) instant
download maybe you interests ...

Computational Intelligence and Intelligent Systems 7th

International Symposium ISICA 2015 Guangzhou China
November 21 22 2015 Revised Selected Papers 1st Edition
Kangshun Li
https://textbookfull.com/product/computational-intelligence-and-
intelligent-systems-7th-international-symposium-
isica-2015-guangzhou-china-november-21-22-2015-revised-selected-
papers-1st-edition-kangshun-li/

Computational Intelligence and Intelligent Systems 10th

International Symposium ISICA 2018 Jiujiang China
October 13 14 2018 Revised Selected Papers Hu Peng

https://textbookfull.com/product/computational-intelligence-and-
intelligent-systems-10th-international-symposium-
isica-2018-jiujiang-china-october-13-14-2018-revised-selected-
papers-hu-peng/

Computational Intelligence and Intelligent Systems

Kangshun Li

https://textbookfull.com/product/computational-intelligence-and-
intelligent-systems-kangshun-li/

Computational Linguistics and Intelligent Text

Processing 18th International Conference CICLing 2017
Budapest Hungary April 17 23 2017 Revised Selected
Papers Part II Alexander Gelbukh
https://textbookfull.com/product/computational-linguistics-and-
intelligent-text-processing-18th-international-conference-
cicling-2017-budapest-hungary-april-17-23-2017-revised-selected-
Neural Information Processing: 24th International
Conference, ICONIP 2017, Guangzhou, China, November
14-18, 2017, Proceedings, Part II 1st Edition Derong
Liu Et Al. (Eds.)
https://textbookfull.com/product/neural-information-
processing-24th-international-conference-iconip-2017-guangzhou-
china-november-14-18-2017-proceedings-part-ii-1st-edition-derong-
liu-et-al-eds/

Image and Graphics 9th International Conference ICIG

2017 Shanghai China September 13 15 2017 Revised
Selected Papers Part III 1st Edition Yao Zhao

https://textbookfull.com/product/image-and-graphics-9th-
international-conference-icig-2017-shanghai-china-
september-13-15-2017-revised-selected-papers-part-iii-1st-
edition-yao-zhao/

Computational Linguistics and Intelligent Text

Processing 18th International Conference CICLing 2017
Budapest Hungary April 17 23 2017 Revised Selected
Papers Part I Alexander Gelbukh
https://textbookfull.com/product/computational-linguistics-and-
intelligent-text-processing-18th-international-conference-
cicling-2017-budapest-hungary-april-17-23-2017-revised-selected-
papers-part-i-alexander-gelbukh/

Image and Video Technology 8th Pacific Rim Symposium

PSIVT 2017 Wuhan China November 20 24 2017 Revised
Selected Papers 1st Edition Manoranjan Paul

https://textbookfull.com/product/image-and-video-technology-8th-
pacific-rim-symposium-psivt-2017-wuhan-china-
november-20-24-2017-revised-selected-papers-1st-edition-
manoranjan-paul/

Studies on Speech Production 11th International Seminar

ISSP 2017 Tianjin China October 16 19 2017 Revised
Selected Papers Qiang Fang

https://textbookfull.com/product/studies-on-speech-
production-11th-international-seminar-issp-2017-tianjin-china-
october-16-19-2017-revised-selected-papers-qiang-fang/
Kangshun Li
Wei Li
Zhangxing Chen
Yong Liu (Eds.)

Communications in Computer and Information Science 874

Computational Intelligence
and Intelligent Systems
9th International Symposium, ISICA 2017
Guangzhou, China, November 18–19, 2017
Revised Selected Papers, Part II

123
Communications
in Computer and Information Science 874
Commenced Publication in 2007
Founding and Former Series Editors:
Phoebe Chen, Alfredo Cuzzocrea, Xiaoyong Du, Orhun Kara, Ting Liu,
Dominik Ślęzak, and Xiaokang Yang

Editorial Board
Simone Diniz Junqueira Barbosa
Pontifical Catholic University of Rio de Janeiro (PUC-Rio),
Rio de Janeiro, Brazil
Joaquim Filipe
Polytechnic Institute of Setúbal, Setúbal, Portugal
Igor Kotenko
St. Petersburg Institute for Informatics and Automation of the Russian
Academy of Sciences, St. Petersburg, Russia
Krishna M. Sivalingam
Indian Institute of Technology Madras, Chennai, India
Takashi Washio
Osaka University, Osaka, Japan
Junsong Yuan
University at Buffalo, The State University of New York, Buffalo, USA
Lizhu Zhou
Tsinghua University, Beijing, China
More information about this series at http://www.springer.com/series/7899
Kangshun Li Wei Li
•

Zhangxing Chen Yong Liu (Eds.)

•

Computational Intelligence
and Intelligent Systems
9th International Symposium, ISICA 2017
Guangzhou, China, November 18–19, 2017
Revised Selected Papers, Part II

123
Editors
Kangshun Li Zhangxing Chen
College of Mathematics and Informatics Chemical and Petroleum Engineering
South China Agricultural University University of Calgary
Guangzhou Calgary, AB
China Canada
Wei Li Yong Liu
Jiangxi University of Science School of Computer Science
and Technology and Engineering
Ganzhou, Jiangxi The University of Aizu
China Aizu-Wakamatsu, Fukushima
Japan

ISSN 1865-0929 ISSN 1865-0937 (electronic)

Communications in Computer and Information Science
ISBN 978-981-13-1650-0 ISBN 978-981-13-1651-7 (eBook)
https://doi.org/10.1007/978-981-13-1651-7

Library of Congress Control Number: 2018948807

© Springer Nature Singapore Pte Ltd. 2018

This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of the
material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation,
broadcasting, reproduction on microfilms or in any other physical way, and transmission or information
storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now
known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication
does not imply, even in the absence of a specific statement, that such names are exempt from the relevant
protective laws and regulations and therefore free for general use.
The publisher, the authors, and the editors are safe to assume that the advice and information in this book are
believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the editors
give a warranty, express or implied, with respect to the material contained herein or for any errors or
omissions that may have been made. The publisher remains neutral with regard to jurisdictional claims in
published maps and institutional affiliations.

This Springer imprint is published by the registered company Springer Nature Singapore Pte Ltd.
The registered company address is: 152 Beach Road, #21-01/04 Gateway East, Singapore 189721,
Singapore
 Preface

Volumes CCIS 873 and CCIS 874 comprise proceedings of the 9th International
Symposium on Intelligence Computation and Applications (ISICA 2017) held in
Guangzhou, China, during November 18–19, 2017. ISICA 2017 successfully attracted
over 180 submissions. After rigorous reviews and plagiarism checking, 51 high-quality
papers are included in CCIS 873, while another 50 papers are collected in CCIS 874.
ISICA conferences are one of the first series of international conferences on compu-
tational intelligence that combines elements of learning, adaptation, evolution, and
fuzzy logic to create programs as alternative solutions to artificial intelligence.
ISICA 2017 featured the most up-to-date research in analysis and theory of evo-
lutionary computation, neural network architectures and learning, neuro-dynamics and
neuro-engineering, fuzzy logic and control, collective intelligence and hybrid systems,
deep learning, knowledge discovery, learning, and reasoning. ISICA 2017 provided a
venue to foster technical exchanges, renew everlasting friendships, and establish new
connections. Prof. Yuanxiang Li, one of the pioneers in parallel and evolution com-
puting at Wuhan University, wrote a beautiful poem in Chinese for the ISICA 2017
event. It is our pleasure to translate his poem with the title of “Computational Intel-
ligence Debate on the Pearl River”:
Wear a smile on a bright face;
Under the night light on the Pearl River;
You are like star and moon shining on the Tower Small Slim Waist;
Ride waves on the cruise ship;
Leave bridges behind in a boundless moment.

You are from far away;

A journey of thousand miles;
Meet in the Guangzhou City;
Brighten up the field of intelligent evolution;
Explore the endless road to intelligence.
Prof. Li’s poem points out one of ISICA’s missions of pursuing the truth that a
complex system inherits the simple mechanism of evolution, while simple models may
lead to the evolution of complex morphologies. Following the success of the past eight
ISICA events, ISICA 2017 continued to explore the new problems emerging in the
fields of computational intelligence.
On behalf of the Organizing Committee, we would like to thank warmly the
sponsors, South China Agricultural University, who helped in one way or another to
achieve our goals for the conference. We wish to express our appreciation to Springer
for publishing the proceedings of ISICA 2017. We also wish to acknowledge the
VI Preface

dedication and commitment of both the staff at the Springer Beijing Office and the
CCIS editorial staff. We would like to thank the authors for submitting their work, as
well as the Program Committee members and reviewers for their enthusiasm, time, and
expertise. The invaluable help of active members from the Organizing Committee,
including Wei Li, Hui Wang, Lei Yang, Yan Chen, Lixia Zhang, Weiguang Chen,
Zhuozhi Liang, Junlin Jin, Ying Feng, and Yunru Lu, in setting up and maintaining the
online submission systems by EasyChair, assigning the papers to the reviewers, and
preparing the camera-ready version of the proceedings is highly appreciated. We would
like to thank them personally for helping to make ISICA 2017 a success.

March 2018 Kangshun Li

Yong Liu
Wei Li
Zhangxing Chen
 Organization

Honorary Chairs
Hisao Ishibuchi Osaka Prefecture University, Japan
Qingfu Zhang City University of Hong Kong, SAR China
Yang Xiang Deakin University, Australia

General Chairs
Kangshun Li South China Agricultural University, China
Zhangxing Chen University of Calgary, Canada
Yong Liu University of Aizu, Japan

Program Chairs
Aniello Castiglione University of Salerno, Italy
Jing Liu Xidian University, China
Han Huang South China University of Technology, China
Hailin Liu Guangdong University of Technology, China

Local Arrangements Chairs

Wei Li South China Agricultural University, China
Yan Chen South China Agricultural University, China

Publicity Chairs
Lei Yang South China Agricultural University, China
Lixia Zhang South China Agricultural University, China

Program Committee
Aimin Zhou East China Normal University, China
Allan Rocha University of Calgary, Canada
Dazhi Jiang Shantou University, China
Dongbo Zhang Guangdong University of Science and Technology,
China
Ehsan Aliabadian University of Calgary, Canada
Ehsan Amirian University of Calgary, Canada
Feng Wang Wuhan University, China
Guangming Lin Southern University of Science and Technology, China
Guoliang He Wuhan University, China
VIII Organization

Hailin Liu Guangdong University of Technology, China

Hu Peng Jiujiang University, China
Hui Wang Nanchang Institute of Technology, China
Iyogun Christopher University of Calgary, Canada
Jiahai Wang Sun Yet-Sen University, China
Jing Wang Jiangxi University of Finance and Economics, China
Jun He Aberystwyth University, UK
Jun Zou The Chinese University of Hong Kong, SAR China
Kangshun Li South China Agricultural University, China
Ke Tang Southern University of Science and Technology, China
Kejun Zhang Zhejiang University, China
Lingling Wang Wuhan University, China
Lixin Ding Wuhan University, China
Lu Xiong South China Agricultural University, China
Maoguo Gong Xidian University, China
Mohammad Zeidani University of Calgary, Canada
Rafael Almeida University of Calgary, Canada
Sanyou Zeng China University of Geosciences, China
Shenwen Wang Shijiazhuang University of Economics, China
Wayne Li University of Calgary, Canada
Wei Li South China Agricultural University, China
Wensheng Zhang Chinese Academy of Sciences, China
Xiangjing Lai University of Angers, France
Xin Du Fujian Normal University, China
Xinyu Zhou Jiangxi Normal University, China
Xuesong Yan China University of Geosciences, China
Xuewen Xia East China Jiaotong University, China
Ying Huang Gannan Normal University, China
Yong Liu The University of Aizu, Japan
Zahra Sahaf University of Calgary, Canada
Zhangxing Chen University of Calgary, Canada
Zhun Fan Shantou University, China
 Contents – Part II

Swarm Intelligence – Cooperative Search

Differential Opposition-Based Particle Swarm . . . . . . . . . . . . . . . . . . . . . . . 3

Lanlan Kang, Wenyong Dong, Shanni Li, and Jianxin Li

Research on Hierarchical Cooperative Algorithm Based on Genetic

Algorithm and Particle Swarm Optimization . . . . . . . . . . . . . . . . . . . . . . . . 16
Linrun Qiu

An Adaptive Particle Swarm Optimization Using Hybrid Strategy . . . . . . . . . 26

Peng Shao, Zhijian Wu, Hu Peng, Yinglong Wang, and Guangquan Li

ITÖ Algorithm with Cooperative Coevolution for Large Scale

Global Optimization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
Yufeng Wang, Wenyong Dong, and Xueshi Dong

A Conical Area Differential Evolution with Dual Populations

for Constrained Optimization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
Bin Wu, Weiqin Ying, Yu Wu, Yuehong Xie, and Zhenyu Wang

Swarm Intelligence – Swarm Optimization

A Particle Swarm Clustering Algorithm Based on Tree Structure

and Neighborhood. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
Lei Yang, Wensheng Zhang, Zhicheng Lai, and Ziyu Cheng

Optimization of UWB Antenna Based on Particle Swarm

Optimization Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
Mingyuan Yu, Jing Liang, Boyang Qu, and Caitong Yue

A Divisive Multi-level Differential Evolution . . . . . . . . . . . . . . . . . . . . . . . 98

Huifang Zhang, Wei Huang, and Jinsong Wang

Complex Systems Modeling – System Dynamic

A Comparative Summary of the Latest Version of MapReduce Parallel

and Old Version from the Perspective of Framework . . . . . . . . . . . . . . . . . . 113
Xinze Li and Qi Liu

A Third-Order Meminductor Chaos Circuit with Complicated Dynamics . . . . 125

Zhiping Tan and Shanni Li
X Contents – Part II

Mathematical Model of Cellular Automata in Urban Taxi

Network – Take GanZhou as an Example . . . . . . . . . . . . . . . . . . . . . . . . . 133
Zhaosheng Wang and Shiyu Li

Hybrid Colliding Bodies Optimization for Solving Emergency Materials

Transshipment Model with Time Window . . . . . . . . . . . . . . . . . . . . . . . . . 142
Xiaopeng Wu, Yongquan Zhou, and Qifang Luo

A Dual Internal Point Filter Algorithm Based on Orthogonal Design . . . . . . . 152

Yijin Yang, Tianyu Huo, Bin Lan, and Sanyou Zeng

Complex Systems Modeling – Multimedia Simulation

A Beam Search Approach Based on Action Space for the 2D Rectangular

Packing Problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165
Aihua Yin, Lei Wang, Dongping Hu, Hao Rao, and Song Deng

On the Innovation of Multimedia Technology to the Management Model

of College Students . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175
Yuanbing Wang

Convenient Top-k Location-Text Publish/Subscribe Scheme . . . . . . . . . . . . . 183

Hong Zhu, Hongbo Li, Zongmin Cui, Zhongsheng Cao, and Meiyi Xie

Effects of Foliar Selenium Fertilizer on Agronomical Traits and Selenium,

Cadmium Contents of Different Rape Varieties. . . . . . . . . . . . . . . . . . . . . . 192
Bin Du, HuoYun Chen, and DanYing Xing

Fresh-Water Fish Quality Traceability System Based on NFC Technology . . . 204

Longqing Zhang, Lei Yang, Liping Bai, Yanghong Zhang,
and Kaiming You

Intelligent Information Systems – Information Retrieval

An Information Filtering Model Based on Neural Network. . . . . . . . . . . . . . 217

Rongrong Li

The Theory of Basic and Applied Research in Information Retrieval

Sorting Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 228
Xinze Li, Jiying Yang, and Qi Liu

Summary of Research on Distribution Centers . . . . . . . . . . . . . . . . . . . . . . 238

Zeping Li and Huwei Liu

Factorization of Odd Integers as Lattice Search Procedure . . . . . . . . . . . . . . 251

Xingbo Wang
 Contents – Part II XI

Research on Key Technology of Distributed Indexing and Retrieval

System Based on Lucene . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 259
Rongrong Li

Intelligent Information Systems – E-commerce Platforms

Research on the Integrated Development Model of e-Commerce

Channel and Physical Retail Channel. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 273
Sisi Li

Study on Potency of Controlling on Crematogaster Rogenhoferi

to Parasaissetia Nigra Nietner . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 280
Lihe Zhang, Bin Du, Baoli Qiu, and Hui Wang

Research on the Management and Optimization of Warehouse

Location in e-Commerce Enterprises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 290
Huwei Liu and Zeping Li

A New SOC Estimation Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 303

Weihua Zhong, Fahui Gu, and Wenxiang Wang

Analysis on Current Situation of E-Commerce Platform

for the Development from C2M Model to C2B Model. . . . . . . . . . . . . . . . . 312
Bo Yang

On the Artistic Characteristics of Computer Aided Design

in Fashion Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 322
Ping Wang

Artificial Intelligence and Robotics – Query Optimization

Rock-Paper-Scissors Game Based on Two-Domain DNA

Strand Displacement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 331
Wendan Xie, Changjun Zhou, Xianwen Fang, Zhixiang Yin,
and Qiang Zhang

A Business Resource Scheduled Algorithm of TD-LTE Trunking

System Based on QoS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 341
Qiutong Li, Yuechen Yang, and Baocai Zhong

Assumption Queries Processing of Probabilistic Relational Databases . . . . . . 354

Caicai Zhang, Zongmin Cui, and Hairong Yu

Design and Implementation of Self-balancing Robot Based on STM32 . . . . . 365

Ling Peng and Chunhui Zhou
XII Contents – Part II

The Design and Implementation of a Route Skyline Query System

Based on Weighted Voronoi Diagrams . . . . . . . . . . . . . . . . . . . . . . . . . . . 376
Jiping Zheng, Yiwei Ding, Shunqing Jiang, and Zhongling He

Improved RFID Anti-collision Algorithm Based on Quad-Tree . . . . . . . . . . . 390

Hui Guan, Zhaobin Liu, and Yan Zhang

Artificial Intelligence and Robotics – Intelligent Engineering

Discussion on the Important Role of Computer-Aided Intelligent

Manufacturing in the Transition of Garment Industry
to Softening Production . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 403
Ping Wang

A Study of Miniaturized Wide-Band Antenna Design . . . . . . . . . . . . . . . . . 413

Rui Zhang, Jianqing Sun, Yongzhi Sun, Bin Lan, and Sanyou Zeng

Yagi-Uda Antenna Design Using Differential Evolution . . . . . . . . . . . . . . . . 427

Hai Zhang, Hui Wang, and Cong Wang

Research on Coordination Fresh Product Supply Chain Under

New Retailing Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 439
Bo Yang and Dongbo Zhang

Virtualization – Motion-Based Tracking

Real-Time RGBD Object Tracking via Collaborative Appearance

and Motion Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 449
Danxian Chen, Zhanming Liu, Hefeng Wu, and Jin Zhan

Lip Password-Based Speaker Verification Without a Priori Knowledge

of Speech Language . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 461
Yiu-ming Cheung and Yichao Zhou

Human Motion Model Construction Based on Gene

Expression Programming . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 473
Wei He, Shaoyang Hu, Shanni Li, Junlin Jin, and Kangshun Li

Research of Crowed Abnormal Behavior Detection Technology Based

on Trajectory Gradient . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 486
Kangshun Li, Hongtao Huang, Zebiao Zheng, and Yusheng Lu

A Novel Monitor Image De-hazing for Heavy Haze on the Freeway . . . . . . . 501
Chunyu Xu, Yufeng Wang, and Wenyong Dong

Real-Time Tracking with Multi-center Kernel Correlation Filter . . . . . . . . . . 512

Taoe Wu, Zhiqiang Zhao, Zongmin Cui, Anyuan Deng, and Xiao Yang
 Contents – Part II XIII

Virtualization – Image Recognition

The Reorganization of Handwritten Figures Based on Convolutional

Neural Network . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 525
Xingzhen Tao, Wenxiang Wang, and Lei Lu

Comparison of Machine Learning Algorithms for Handwritten

Digit Recognition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 532
Shixiao Wu, Wanyun Wei, and Libing Zhang

A User Identification Algorithm for High-Speed Rail Network Based

on Switching Link. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 543
Wenxiang Wang and Xingzhen Tao

A New Language Evolution Model for Chinese Spatial Preposition . . . . . . . . 551

Qi Rao and Youjie Zheng

Research on Location Technology Based on Mobile Reference Nodes . . . . . . 561

Xuefeng Yang, Lin Li, and Yue Liu

Author Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 571

 Contents – Part I

Neural Networks and Statistical Learning – Neural

Architecture Search

A New Recurrent Neural Network with Fewer Neurons for Quadratic

Programming Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
Sanfeng Chen, Xin Han, Fei Tang, and Guangming Lin

Mutual-Information-SMOTE: A Cost-Free Learning Method

for Imbalanced Data Classification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
Ying Chen, Yufei Chen, Xianhui Liu, and Weidong Zhao

Ontology Sparse Vector Learning Algorithm . . . . . . . . . . . . . . . . . . . . . . . 31

Xin Xin Huang and Shu Gong

Bacterial Foraging Algorithm Based on Reinforcement Learning

for Continuous Optimizations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
Huiyan Jiang, Wanpeng Dong, Lianbo Ma, and Rui Wang

A Novel Attribute Reduction Approach Based on Improved

Attribute Significance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
Jun Ye and Lei Wang

Neural Networks and Statistical Learning – Transfer of Knowledge

Traffic Condition Assessment Based on Support Vectors Machine

Using Intelligent Transportation System Data . . . . . . . . . . . . . . . . . . . . . . . 69
Deng Lei and Weihua Zhong

Bidirectional Negative Correlation Learning . . . . . . . . . . . . . . . . . . . . . . . . 84

Yong Liu

Reflectance Estimation Based on Locally Weighted Linear

Regression Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
Dejun Lu, Weifeng Zhang, Kaixuan Cuan, and Pengfei Liu

A Multi-task Learning Approach for Mandarin-English Code-Switching

Conversational Speech Recognition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
Xiao Song, Yi Liu, Daming Yang, and Yuexian Zou

Feature Selection of Network Flow Based on Machine Learning . . . . . . . . . . 112

Taian Xu
XVI Contents – Part I

Evolutionary Multi-objective and Dynamic Optimization

– Optimal Control and Design

Multi-objective Optimal Scheduling of Valves and Hydrants for Sudden

Drinking Water Pollution Incident . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
Chengyu Hu, Lu Zou, Xuesong Yan, and Wenyin Gong

A Novel Mutation and Crossover Operator for Multi-objective

Differential Evolution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138
Qingxia Li and Wenhong Wei

Multi-objective Gene Expression Programming Based Automatic

Clustering Method. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148
Ruochen Liu, Jianxia Li, and Manman He

Multi-objective Firefly Algorithm Guided by Elite Particle . . . . . . . . . . . . . . 159

Jiayuan Wang, Li Lv, Zhifeng Xie, Xi Zhang, Hui Wang, and Jia Zhao

Improving Energy Demand Estimation Using an Adaptive Firefly

Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 171
Hui Wang, Zhangxin Chen, Wenjun Wang, Zhijian Wu, Keliu Wu,
and Wei Li

Evolutionary Multi-objective and Dynamic Optimization

– Hybrid Methods

Firefly Algorithm with Elite Attraction . . . . . . . . . . . . . . . . . . . . . . . . . . . 185

Jing Wang

A Hybrid Fireworks Explosion Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . 195

Liping Wang, Renwen Chen, and Chengwang Xie

An Improved Multi-objective Fireworks Algorithm . . . . . . . . . . . . . . . . . . . 204

Dongming Zhan and Chengwang Xie

Evolutionary Design of a Crooked-Wire Antenna . . . . . . . . . . . . . . . . . . . . 219

Lumin Ye, Bin Lan, Yi Yuan, Jianqing Sun, Yongzhi Sun,
and Sanyou Zeng

Typical Constrained Optimization Formulation in Evolutionary

Computation Not Suitable for Expensive Optimization. . . . . . . . . . . . . . . . . 232
Sanyou Zeng, Ruwang Jiao, Changhe Li, Bin Lan, Huanhuan Li,
Jianqing Sun, and Yongzhi Sun
 Contents – Part I XVII

Data Mining – Association Rule Learning

Maize Gene Regulatory Relationship Mining Using Association Rule . . . . . . 249

Jianxiao Liu, Chaoyang Wang, Haijun Liu, Yingjie Xiao, Songlin Hao,
Xiaolong Zhang, Jianchao Sun, and Huan Yu

Database Reengineering Scheme from Object-Oriented Model

to Flattened XML Data Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 259
Yue Liu and Xukun Wu

A Modified Shuffled Frog Leaping Algorithm for Constructing

DNA Codes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 269
Zhenghui Liu, Bin Wang, Changjun Zhou, Xiaopeng Wei,
and Qiang Zhang

Clustering Based Prediction of Financial Data by ARMA Model . . . . . . . . . 279

Duobiao Ning, Siyu Zhang, Wenfei Chen, and Xinqiao Yu

Research on Data Mining Algorithm of Association Rules Based

on Hadoop . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 288
Linrun Qiu

Data Mining – Data Management Platforms

Data-Driven Phone Selection for Language Identification

via Bidirectional Long Short-Term Memory Modeling . . . . . . . . . . . . . . . . . 301
Xiao Song, Qiang Cheng, Jingping Xing, and Yuexian Zou

Multi-document Summarization via LDA and Density Peaks Based

Sentence-Level Clustering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 313
Baoyan Wang, Yuexian Zou, Jian Zhang, Jun Jiang, and Yi Liu

The Dynamic Relationship Between Bank Credit and Real Estate Price
in China. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 324
Xiaofan Wang and Li Zhou

Big-Data Cloud Services Platform for Growth Enterprises with Adaptive

Exception Handling and Parallelized Data Mining . . . . . . . . . . . . . . . . . . . . 339
Yazhi Wen, Hu Bo, and Bin Wen

Research on Automatic Generation of Test Cases Based

on Genetic Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 351
Lu Xiong and Kangshun Li

A Routing Acceleration Strategy via Named Data Networking

in Space-Terrestrial Integrated Networks . . . . . . . . . . . . . . . . . . . . . . . . . . 361
Feng Yang and Di Liu
XVIII Contents – Part I

Cloud Computing and Multiagent Systems – Service Models

Exploring Migration Issue Based on Multi-agent Modeling . . . . . . . . . . . . . 373

Pengfei Liu, Xiaxu He, Weifeng Zhang, and Enkai Chen

Application of Plant-Derived Anti Repellents in Prevention and Cure

of Parasaissetia Nigra Nietner . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 383
Lihe Zhang, Bin Du, Baoli Qiu, and Hui Wang

Research on Resource Trust Access Control Based on Cloud

Computing Environment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 394
Jun Nie and Dongbo Zhang

A Statistical Study of Technological Innovation Factors in Beijing’s

Low-Carbon Economic Growth. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 405
Xiaofan Wang and Li Zhou

Cloud Computing and Multiagent Systems – Cloud Engineering

Study on Critical Lines Identification in Complex Power Grids. . . . . . . . . . . 423

Yi Wang, Zhiping Tan, and Yanli Zou

A Review of Multi-sensor Data Fusion for Traffic. . . . . . . . . . . . . . . . . . . . 432

Xue Zhao and Dongbo Zhang

Research on Evaluation Method of Service Quality . . . . . . . . . . . . . . . . . . . 445

Yan Zhao, Xiaxia Niu, and Li Zhou

Research on Traffic Data Fusion Based on Multi Detector . . . . . . . . . . . . . . 456

Suping Liu

A Recommend Method of Hotspots Knowledge Based on Big Data

from Evolving Network . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 468
Yi Zhao, Zhao Li, and Jun Wu

Everywhere Connectivity – IoT Solutions

Improved Location Algorithm Based on DV-Hop for Indoor Internet

of Things . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 483
Qian Cai

A Design of the Shared Farmland System Based on the Internet

of Things Technology and IMS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 492
Na Chang and Junhua Ku

Personalized Recommendation Algorithm Based on Commodity Label . . . . . 499

Yuehua Dong and Xuelei Liang
 Contents – Part I XIX

Mobile Node Localization Based on Angle Self-adjustment with Mine

Wireless Sensor Networks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 513
Wangsheng Fang, Hui Wang, and Zhongdong Hu

Research on the Development Strategy of O2O e-Commerce in Traditional

Retail Enterprises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 521
Sisi Li and Suping Liu

Implementation of Academic News Recommendation System Based

on User Profile and Message Semantics . . . . . . . . . . . . . . . . . . . . . . . . . . . 531
Weiling Li, Yong Tang, Guohua Chen, Danyang Xiao,
and Chengzhe Yuan

Everywhere Connectivity – Wireless Sensor Networks

An Optimal Sink Placement for High Coverage and Low Deployment

Cost in Mobile Wireless Sensor Networks . . . . . . . . . . . . . . . . . . . . . . . . . 543
Qingzhong Liang and Yuanyuan Fan

Wireless Sensor Network Time Synchronization Algorithm Overview . . . . . . 552

Chunqiang Liu, Haijie Pang, Ning Cao, Xinze Li, and Dongchen Xu

Collaborative Filtering Recommendation Considering Seed Life Cycle

for BT Download Websites . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 562
Yue Liu, Fei Cai, Qi Sun, and Yiming Zhu

Centralized Access Control Scheme Based on OAuth for Social Networks . . . 576
Yue Liu, Wei Gao, and Jingyun Liao

Research on Localization Scheme of Wireless Sensor Networks Based

on TDOA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 588
Xuefeng Yang, Junqi Ma, and Yuting Lu

Author Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 601

Swarm Intelligence – Cooperative
Search
 Differential Opposition-Based Particle Swarm

Lanlan Kang1(&), Wenyong Dong2, Shanni Li3, and Jianxin Li4

1
Jiangxi University of Science and Technology, Ganzhou 341000, China
victorykll@163.com
2
Wuhan University, Wuhan 430072, China
3
Southern Capital Management Co., Ltd., Shenzhen 518000, China
4
Dongguan Polytechnic, Dongguan 523808, Guangdong, China

Abstract. Particle Swarm Optimization (PSO) is slow but steady learner

although it exhibits strong competence in solving complicated problems.
However, during the course of searching process, the particles gradually gather
into the vicinity of the best particle found so far. Furthermore, some evidences
show that the unreasonable setting of its inertial term in the kinetic equations
may lead to slow convergence of PSO. Thus, a differential opposition-based
particle swarm optimization with adaptive elite mutation (DOPSO) is presented
to overcome these drawbacks in this paper. There are two strategies are intro-
duced into DOPSO to balance the contradiction between exploration and
exploitation during its searching process: (1) Firstly a new particle’s position
update rule in which differential term replaces the inertia term is designed to
accelerate its convergence; (2) Secondly an adaptive elite mutation strategy
(AEM) is included to avoid trapping into local optimum. Experimental results
show that the proposed method has a significant improvement in performance
compared with some state-of-art PSOs.

Keywords: Particle swarm optimization

Generalized opposition-based learning
Adaptive elite mutation
Differential term

1 Introduction

Particle swarm optimization (PSO), one of the popular Evolutionary algorithm

(EA) [1], firstly proposed by Kennedy and Eberhart in 1995. Inspired by the simulation
of bird flocks and fish schooling foraging behavior, particles in PSO could search for
global optimum through cooperation between particles [2]. At present, PSO has been
successfully applied to a number of applications, such as resource allocation [3],
prognostic model design for metabolic syndrome [4], DNA sequence compression [5]
and so on. However, it can be found that the more complex problems, the slower PSO
perform and the worse search in efficiency. As a result, the weakness can cause per-
formance degradation of PSO, such as overmuch computation overhead, premature
convergence.
There are many strategies to improve the performance of basic PSO, such as
parameters adaptation tricks, opposition-based searching methods and hybridization of
PSO with other intelligent algorithm.

© Springer Nature Singapore Pte Ltd. 2018

K. Li et al. (Eds.): ISICA 2017, CCIS 874, pp. 3–15, 2018.
https://doi.org/10.1007/978-981-13-1651-7_1
4 L. Kang et al.

Because the performance of basic PSO highly relies on its parameters, such as
inertial term, social learning coefficient, some strategies have been proposed to adap-
tively adjust these parameters for different problem [8]. Zhang et al. [9] employ the
evolutionary state estimation to identify the evolutionary states of the swarm as
exploration, exploitation, convergence and jumping out, and propose and adaptive
control mechanism to change inertial weight and acceleration coefficients. Hu et al. [10]
further put forward a parameter control mechanism to adaptively adjust parameters at
iteration by subgradient method to close to the global best position and thus improve
the robustness of PSO-MAM.
In order to enhance local search ability, Opposition-based learning (OBL) was
firstly introduced into PSO [11]. The empirical researches prove that OBL is effective
strategy to enhance local search ability and increase convergence rate. General-
ized OBL (GOBL) was developed on the basis of OBL in literature [12].
The hybridization of PSO with other intelligent algorithm has been proved to be a
promising technique that combines the desirable properties of these methods to improve
individual weakness. For example, A hybrid operation combining DE and PSO is
devised, where each individual is sequentially carried out DE and PSO operators in
order to make each subpopulation quickly track the moving peaks [13]. Multi-frequency
vibrational PSO employ periodic mutation application strategy and diversity variety
based on an artificial neural network to avoid premature convergence [14].
For purpose of performance, a differential opposition-based particle swarm opti-
mization with adaptive elite mutation, termed as DOPSO, is proposed in this paper. The
algorithm is composed of a novelty velocity update formula (VD), GOBL and AEM
strategy. VD formula differs from traditional PSO, contains a new momentum com-
ponent instead of inertial motion component in basic PSO. The new velocity update
formula can efficiently accelerate the convergence rate and enhance exploitation ability
of algorithm by more widely learning from other individuals. With the introduction of
generalized opposition-based Learning (GOBL) strategy in this paper, an adaptive elite
mutation strategy (AEM) is applied to avoid particles, especially elite particles, trap-
ping into local optimum [15]. Experiment results show that DOPSO has significantly
improved convergence speed compared with the other OBL-based PSOs.
The rest of this paper is organized as follows. Section 2 briefly introduces the basic
PSO algorithm and GOBL strategy. VD formula and AEM strategy are presented in
Sect. 3 including detailed analysis of thought originated and search behaviors. Ulti-
mately, Sect. 3 gives detailed procedure of DOPSO. Section 4 experimentally com-
pares the DOPSO with various opposition-based learning PSO algorithms on 13
benchmark problems. Furthermore, the optimal parameter settings are suggested.
Finally, some conclusions and discussions on the future work are given in Sect. 5.

2 Related Work

2.1 The Basic PSO

PSO is modeled on an abstract framework of “collective intelligence” in social animals.
Each particle of a swarm represents a candidate solution, has two basic properties, i.e.
 Differential Opposition-Based Particle Swarm 5

velocity and position [16]. It is supposed that velocity vi;j and position xi;j of the jth
dimension of the ith particle are updated according to the following equations:

vi;j ðt þ 1Þ ¼ w
vi;j ðtÞ þ c1 rand1 ðpbesti;j  xi;j ðtÞÞ þ c2 rand2 ðgbestj  xi;j ðtÞÞ ð1Þ

xi;j ðt þ 1Þ ¼ xi;j ðtÞ þ vi;j ðt þ 1Þ ð2Þ

Where, i ¼ 1; 2;


; N; j ¼ 1; 2;


; D, N is the size of swarm, pbest is particle’s

historical best position, gbest is the swarm’s global best position, c1 and c2 are two
acceleration coefficients which control the influence of the cognitive and social com-
ponents and guide each particle toward the pbest and the gbest, respectively. w is
inertial weight. In general, c1 ; c2 2 ½0; 2 : rand1 and rand2 are two uniformly dis-
tributed random variables in the interval [0, 1].

2.2 A Generalized Opposition-Based Learning (GOBL)

A generalized opposition-based learning, called GOBL [17]. GOBL applied to PSO is
defined as follows:
Let Xi ¼ ðxi;1 ; xi;2 ;


; xi;D Þ is the ith particle in a D-dimension space. The opposite
^ ^ ^ ^
particle X i ¼ ðxi;1 ; xi;2 ;


; xi;D Þ is defined by:
^
x i;j ¼ k
ðdaj þ dbj Þ  xi;j ð3Þ

Where, j ¼ 1; 2;


; D, k is a random number drawn from the uniform distribution

in the interval [0, 1]. daj and dbj are the minimum and maximum values of the jth
dimension of the ith particle, respectively:

daj ¼ minðxi;j Þ; dbj ¼ maxðxi;j Þ ð4Þ

3 DOPSO Algorithm

This paper proposes DOPSO algorithm, which incorporates a new velocity formula and
two strategies to enhance the global search ability as well as accelerate the convergence
rate of DOPSO. First, GOBL strategy is introduced to generate initial population and
select superior particles in each generation by given probability. Second, we adopt a
new velocity formula that contains no inertial component to generate the next particle
positions. Third, AEM strategy is designed to augment search space and avoid trapping
into local optimum.
In the following sections, the differential velocity formula (VD) will firstly be
elaborated, including its original intention, design principle and formula expression.
Then, AEM strategy will be introduced in detail, including its design motivations and
purposes. Finally, the detailed design steps of DOPSO algorithm are given.
6 L. Kang et al.

3.1 A New Update Equation of Velocity

There is broad recognition that evolution is a slow learner, but it is steady increase in
computing power and is able to find the optimum in the amount of practical problems if
computation budget is sufficient. Therefore, how to increase convergence speed under
the condition of ensuring the accuracy of solutions is to be worth researching. Next,
there is a thorough analysis about PSO for its high search efficiency than most of EAs.
As mentioned, each particle in swarm has a velocity and position. The flight
direction of each particle is modified in search space depend on its velocity formula
constantly updated in each iteration (see Eq. (1)). The velocity formula is composed of
three components, i.e. inertial component, cognitive component and social component.
Although the inertial term in Eq. (1) can bring in diversity of swarm, it can also
lead to slow convergence. Can we find some new strategies to take the role of inertial
term while enhance the performance of PSO?
To answer the above questions, the velocity formula of PSO is depicted as shown in
Fig. 1(a) for Eq. (1). We will divide all the terms in Eq. (1) into two parts: (1) the first
part (shown as red-dotted box) stands for the self-related information part which is
composed of the inertial component and the cognition component; and (2) the second
part stands for the social cognition part which is utilized only by gbest in social
component (shown in blue-dotted box). The illustration shows particles may not
obtained sufficient information from current environment or global information may
not be fully used to lead particles’ flight in the next time.
To address above question, inspired by differential evolution, the velocity formula
is modified using the difference between two individuals randomly sampled in swarm
in Eq. (5), called VD.

vdi;j ðt þ 1Þ ¼ s
ðxr1;j ðtÞ  xr2;j ðtÞÞ þ c1 rand1 ðpbesti;j  xi;j ðtÞÞ þ c2 rand2 ðgbestj  xi;j ðtÞÞ ð5Þ

Where, i ¼ 1; 2;


; N; j ¼ 1; 2;


; D, s is a differential factor, used to control the

scope of the search. r1 and r2 are two different integers selected from a uniform
distribution in the interval [1, N], r1 6¼ r2 6¼ i. Other parameters setting, such as c1 , c2 ,
are the same as Eq. (1).

Self-related Information
( pbesti , j − xi , j (t ))
Cognition component c1 rand1
vi , j (t )
ω
Inertial motion s
xr1, j − xr 2, j
⇓

( pbesti , j − xi , j (t )) c1 rand1
⇓

Momentum component vd i , j (t + 1)
vi , j (t + 1)
Cognition component c2 rand 2
( gbest j − xi , j (t ))
c2 rand 2 Social component
( gbest j − xi , j (t ))
Social component population-related Information

(a) (b)

Fig. 1. Decomposition diagram of update equation of velocity (Color figure online)

 Differential Opposition-Based Particle Swarm 7

Obviously, the first part is displaced by a differential mutation operator inspired by

differential evolution scheme which further strengthen the search abilities by learn from
a pair of particles randomly sampled in swarm, and thus it is called momentum
component and not inertial motion. The structure of new formula shows that particles
have transformed the thinking mode and depends on not only self-experience but also
others experience when updating fly direction in the next time. The momentum
component mostly focuses on population information instead of self-related informa-
tion (as shown in red-dotted box of Fig. 1(b)). Improved formula (VD) could augment
the search range of PSO and produce some additional exploration ability of the search
space by adaptively changing flying direction with the behavior of the swarm, and then
efficiently improve the speed of convergence and calculation accuracy.

3.2 Adaptive Elite Mutation Selection Strategy

In the process of evolution, capacity of searching space of each particle is not the same. In
order to full motivate particles’ activity, “elite strategy” is adopted to speed up the con-
vergence through more evolution is carried out by elite particles whose fitness values in
swarm is better. In this paper, global optimum (gbest) is regarded as elite particle. Through
adaptive mutation operation for gbest in every generation, a new particle is generated for
further exploration in search space. If the fitness of it is better than original gbest, the new
particle will substitute for gbest and then enters into the next generation. Above thought of
elite mutation is called adaptive elite mutation strategy, shortly AEM [15].
In AEM, generated factor xm for mutation in the ith dimension is defined by:

xmðiÞ ¼ exp ðk
t=tmax Þ
ð1  r ðiÞ=rmax Þ ð6Þ

Where, k is a constant, t is iteration times, tmax is the maximum of t, rðiÞ is the

distance from particles to gbest in ith dimension in a generation. rmax is the maximum
of r vector. rðiÞ is defined as follows:

rðiÞ ¼ jgbestðiÞ  avg pbestðiÞj ð7Þ

Where, i ¼ 1; 2;


; D, avg pbestðiÞ, a mean value of the pbest of N particles, is

formulated as follows:
!,
X
N
avg pbestðiÞ ¼ pbest½j½i N ð8Þ
j¼1

Where, pbest½ j½i is the value of pbest in the ith dimension of the jth population. N is
the size of population.
xm as a generated factor inserts into mutation function (9):
 
1
F ðxmÞ ¼ sign
arctanðxmÞ þ C ð9Þ
p
8 L. Kang et al.

Where, symbol variable sign 2 f1; 1g, C is an undetermined constant which will
get different value as Eq. (9) according to clustering situation of population.
8
< 1:5; st d\102
C ¼ 1:0; 10  st d\101
2 ð10Þ
:
0:5; other

In Eq. (10), st_d which means standard deviation of fitness is defined as Eq. (11) in
this paper and used to measure the clustering degree of individuals. The value of st_d is
divided into three sections: ð0; 102 , ð102 ; 101  and ð101 ; þ 1Þ (the distribution of
fitness value within segmentation points is studied in analysis of strategy).

N 
X 
 fi  fgbest 
st d ¼  f  ð11Þ
i¼1 gbest

Where, fi is the fitness of the ith individual, fgbest is the fitness of current global best
position. Equation (11) shows st d is less when all particles are closer to the gbest, and
vice versa.
According to the above mentioned analysis, an adaptive elite mutation strategy
(AEM) is defined as follows:

gbest ¼ gbest þ FðxmÞ ð12Þ

In each generation, recorded the position of the ith dimension after mutation
operation using AEM, the global best mutation position is be generated, that is gbest  .

3.3 DOPSO Algorithm

A new opposition-based PSO inspired by differential thought, called DOPSO, is
composed of differential velocity update formula (VD) and adaptive elite mutation
(AEM) applied to opposition-based learning PSO. The main steps of DOPSO algorithm
are shown as Table 1. Where jr is the probability of using GOBL operator.

4 Experiments

In this section, two parts of experiment are carried out based on a set of benchmark
problems. The one part is performance comparison, the performance of DOPSO is
compared with a set of OBL-based PSOs. The second part is parameter sensitivity
study, some suggestions about parameter settings are given via sensitivity analysis of
key parameters.

4.1 Benchmark Problems

All experiments are conducted by MATLAB2012 on 13 benchmark problems with
dimension D ¼ 30 and the size of population N ¼ 40. According to their properties,
 Differential Opposition-Based Particle Swarm 9

Table 1. The main steps of DOPSO algorithm.

1 randomly initializes N particles in the population P;

2 generate the opposite solutions of N particles to construct opposite population OP according to
Eq.(5);
3 calculate the fitness of P and OP;
4 select N best fittest solutions from P U OP as an initial population P;
5 update the fitness of P;
6 while the stopping criterion is not meet do
7 If rand(0,1)<jr then
8 update the dynamic interval boundaries [da j , db j ] according to Eq.(5);
9 generate the new opposite population OP of N particles according to Eq.(4);
10 calculate the fitness of OP ;
11 select N best fittest solutions from P U OP as a new current population P;
12 update pbest vector and gbest if needed;
13 else
14 for i=1 to N do
15 calculate the velocity of the ith particle according to Eq.(6) ;
16 update the position of the ith particle according to Eq.(2);
17 update the fitness the ith particle;
18 update pbesti if needed;
19 end for;
20 update gbest if needed;
21 end if
22 for j=1 to D do
23 generate gbest * via mutation operation Eq.(13);
24 end for
25 if the fitness of gbest * is better than gbest
gbest = gbest *
26 end if
27 end while

the problems are divided into two categories, which are unimodal problems f1 f7 and
multimodal problems f7 f13 . Compared with unimodal, multimodal problem is difficult
since the number of local optima grows exponentially with the increase of dimen-
sionality. All test functions are to be minimized in the following experiments. A brief
description of these benchmark problems is listed in Table 2.
Noted, in rotated functions f11 f13 and f7 , M is an D  D orthogonal matrix, and
x ¼ ðx1 ; x2 ;


; xD Þ is a D-dimensional row vector.

4.2 Parameter Settings

The selection of the parameters which is very important to bionic algorithm can greatly
influence the performance of PSO. In comparison between algorithms section, the
parameter settings in this paper are consistent with the original algorithms as far as
possible so as to better compare the performance of DOPSO with other similar
10 L. Kang et al.

Table 2. The 13 benchmark functions in the experiments, where D is the dimension, fmin is the
global minimum value of the test function.
Test function D Search space fmin Function
PD D
Unimodal f1 ðxÞ ¼ i¼1 xi
2 30 [−100, 100] 0 Sphere
PD
f2 ðxÞ ¼ I¼1 ½ðxi þ 0:5Þ
2 30 [−100, 100]D 0 Step
PD1
f3 ðxÞ ¼ i¼1 ½100
ðxi þ 1  x2i Þ2 2
þ ð1  xi Þ  30 [−30, 30]D 0 Rosenbrock
PD Pi 2 D
f4 ðxÞ ¼ i¼1 ð j¼1 xj Þ
30 [−100, 100] 0 Quadric
PD QD
f5 ðxÞ ¼ i¼1 jxi j i¼1 jxi j
30 [−10, 10]D 0 Schwefel2.22
PD i1
30 [−100, 100]D 0 Elliptic
f6 ðxÞ ¼ i¼1 ð10 Þ 6 D1 x2i
f7 ðxÞ ¼ f6 ðzÞ; z ¼ x  M 30 [−5.12, 0 Rotated elliptic
5.12]D
P
Multimodal f8 ðxÞ ¼ D i¼1 ½xi  10 cosð2pxi Þ þ 10
2 30 [−5.12, 0 Rastrigin
5.12]D
 rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
XD . ffi 30 [−32, 32]D 0 Ackley
f9 ðxÞ ¼ 20
exp 0:2
i¼1 i
x 2 D

XD . 
 exp i¼1
cosð2pxi Þ D þ 20 þ e
P QD  
f10 ðxÞ ¼ 4000 D xiffi 30 [−600, 600]D 0 Griewank
i¼1 xi  þ1
1 2 p
i¼1 cos i
f11 ðxÞ ¼ f8 ðzÞ; z¼xM 30 [−32, 32]D
0 Rotated
Rastrigin
f12 ðxÞ ¼ f9 ðzÞ; z ¼ x  M 30 [−600, 600]D 0 Rotated Ackley
f13 ðxÞ ¼ f10 ðzÞ; z ¼ x  M 30 [−32, 32]D 0 Rotated
Griewank

Table 3. The specific parameter settings of DOPSO.

c1 c2 s jr k N D
1.49618 1.49618 0.2 0.3 10 40 30

algorithms. The velocity v is limited to the half range of the search space on each
dimension. Experiments are always performed in running 30 times. The default
parameter settings of DOPSO are listed as Table 3.

4.3 Performance Comparison Between OBL-Based PSO

The performance of DOPSO is compared with OBL-based PSOs including EOPSO
[18], GOPSO [12], OVCPSO [19] and OPSO [11] in this paper. Every algorithm runs
10000 generations every time, all the experiments were conducted 30 times and then
recorded the mean value of the global best optimum of each algorithm as Table 4.
Where Symbol “+” denotes the performance of DOPSO is better than other algorithm,
symbol “−” denotes worse and “*” denotes equal to other algorithm. The best results
in six algorithms aim at the thirteen benchmark problems are shown in bold.
It can be concluded from Table 4 that DOPSO has almost achieved good results in
all benchmark problems. Experimental results show that DOPSO outperforms all test
 Differential Opposition-Based Particle Swarm 11

Table 4. Mean value of the global optimum in 30 runs among six OBL-based PSO aiming at
the 13 benchmark functions.
Funs. OPSO GOPSO OVCPSO EOPSO DOPSO
Multimodal f1 4.59E−36 + 0.00E+00 * 5.00E−01 + 0.00E+00 * 0.00E
+00
f2 2.09E−35 + 3.02E−321 + 6.67E−2 + 1.97E + 0.00E
−323 +00
f3 7.18E+00 − 2.82E+01 + 8.70E+01 + 1.57E−24 − 5.30E
+00
f4 4.13E+04 + 1.32E+04 + 2.94E+01 + 5.01E−03 + 0.00E
+00
f5 −6.51E + −6.98E + −2.49E + 3.01E + 0.00E
−12 −162 +00 −162 +00
f6 1.43E−32 + 4.21E−316 + 1.39E−04 + 9.88E + 0.00E
−324 +00
f7 9.82E+06 − 1.96E+06 + 3.06E+00 + 5.08E+02 + 0.00E
+00
f8 1.51E+01 + 0.00E+00 * 7.57E−01 + 2.09E+00 + 0.00E
+00
f9 1.85E+00 + 0.00E+00 * 9.99E−01 + 1.86E−01 + 0.00E
+00
f10 3.83E−01 + 0.00E+00 * 2.05E−02 + 1.26E−02 + 0.00E
+00
f11 1.51E+01 + 0.00E+00 * 4.14E+01 + 4.11E+00 + 0.00E
+00
f12 2.98E+00 + 0.00E+00 * 1.97E+00 + 3.41E−01 + 0.00E
+00
f13 2.33E−02 + 0.00E+00 * 6.14E−01 + 1.22E−02 + 0.00E
+00
Total + 12 6 13 10 –
* 0 7 0 1 –
− 1 0 0 1 –

problems than OVCPSO and obviously superior to OPSO and PSO besides f3 . Con-
versely, optimal value of f3 is obtained by EOPSO. It is worth noting that GOPSO, like
DOPSO, obtained the optimal value in seven test functions and near-optimal solution in
two test functions (i.e. f2 and f6). However, these two algorithms are all trapped into
local optima at Rosenbrock function (f3).
Further experiment was conducted to compare the comprehensive performance
between DOPSO and GOPSO. Average number of fitness calls were shown in Fig. 2
after 30 times running DOPSO and GOPSO on the condition that precision is 1016
and maximum iterations is 10000, a striking result in Fig. 2 is that average number of
function calls of DOPSO less significantly than GOPSO in almost test functions, which
proves that DOPSO can quickly converge to optimal value.
Another random document with
no related content on Scribd:
 Fig. 268.—Hyocrinus bethellianus. × 2. (From Wyville Thomson.)

Fig. 269.—Tegmen of Hyocrinus, viewed from above after removal of the arms, ×
8. (From Wyville Thomson.)

Fam. 2. Rhizocrinidae.—Stem long and persistent; cirri confined to

a few near the root or replaced by rooting branches of the stem.
Stem-ossicles thin and pentagonal at the summit, but lower down
compressed, elliptical in section, and united by two ligaments
separated by a transverse ridge for articulation. Patina composed of
exposed basals and a ring of five short radials. Orals large or
vestigial. Covering plates and interradials present. Two genera, both
from great depths in the Atlantic—Rhizocrinus (Fig. 270), with five
arms and well developed orals (attachment by branching root-cirri);
and Bathycrinus, with ten arms and vestigial orals; attachment by
root-like branches of stem (this is essentially the same as root-cirri).

Fig. 270.—Rhizocrinus lofotensis. × 1½. (From Wyville Thomson.)

Fam. 3. Pentacrinidae.—Stem consisting of ossicles which are

pentagonal in section, united in pairs by syzygy, the upper one of
each pair bearing a whorl of cirri and united by five bundles of fibres
of petal-like section with the lower one of the pair above it. No
rooting processes. Patina consists almost entirely of columns of
radials, the basals being almost or completely hidden. Orals absent,
but side-plates in the ambulacral grooves. Two recent genera,
Pentacrinus (Isocrinus), with three radials in each column;
Metacrinus, with five to eight radials in a column, but the third radial
bears a pinnule. Pentacrinus is found in both the Caribbean Sea and
the Pacific Ocean; Metacrinus in the Pacific. It appears that the
Pentacrinidae when young are attached by a foot-plate at the apex
of the stem; but when adult, the stem is broken in two and the
animals, like Antedon, swim by movements of the arms, dragging a
large part of the stem after them, by which they effect temporary
attachment. As in other stalked forms, the cavities of the chambered
organ are prolonged into canals which traverse the stalk; but in this
family there is the peculiarity that a repetition of the chambered
organ is found opposite every whorl of cirri.

Fig. 271.—Pentacrinus asteria. × ¼. (From Wyville Thomson.)

Fam. 4. Holopodidae.—Stem represented by a leathery

noncalcified outgrowth from the base of the calyx; one circle of
radials indistinguishably fused with the basals and with each other to
form the walls of the calyx. Large oral plates, ten short arms. One
genus, Holopus, in shallow water in the Caribbean Sea.

Fig. 272.—Arms and portion of stem of Pentacrinus maclearanus, slightly

enlarged. In this species the basals can be seen. (From Wyville Thomson.)

Fig. 273.—Calyx of Actinocrinus, one of the Camerata, broken open to show

structure. amb, Ambulacral groove enclosed in covering plates; B, basal;
R1, R2, R3, the three radials of a column. (After Zittel.).
Fam. 5. Comatulidae.—Stem in the adult broken off, leaving only a
stump, the centro-dorsal, covered with cirri. Six genera. Antedon (=
Comatula) has already been described; many tropical species have
numerous arms and often side-plates and covering plates.
Actinometra is distinguished by its excentric mouth, and by the fact
that the centro-dorsal is flat and has cirri only round its edges;
Atelecrinus has an acorn-shaped centro-dorsal, and the basals are
externally visible; Eudiocrinus differs from Antedon only in having
five arms; Promachocrinus is a remarkable form, having ten radii
(this is a unique feature in Crinoidea); finally, Thaumatocrinus has
basals externally visible, large persistent orals and interradial plates,
and in addition a short free appendage of several plates on the anal
interradius. Antedon and Actinometra are almost world-wide. Six
species of the first have been recorded from British waters, of which
the commonest is Antedon rosacea; four others are distinguished by
having longer cirri, and do not seem to be well defined; but A.
eschrichtii, a northern form, is larger, and is distinguished by having
long proximal pinnules. The other genera are rare, and occur in deep
water.

When we turn to survey fossil Crinoidea, we are met with a

bewildering variety of forms ranging from the Lower Cambrian to the
present day. As already mentioned, there is no agreement amongst
experts as to how they should be classified. Bather makes the
fundamental cleavage depend on the possession of two whorls of
plates in the base (Fig. 274), or of only one whorl. These two
divisions he calls Dicyclica and Monocyclica respectively. He
admits that in many forms allied to Dicyclica the infra-basals have
disappeared; these he terms "pseudomonocyclic" forms, and
believes that he is able to discriminate them from true Monocyclica.
 Fig. 274.—Crotalocrinus pulcher. × 1. B, basal; Br, arm-fan of adhering
branches; col, ossicle of stem; IB, infra-basal; R, radial. (After Zittel.)

The present author is utterly unable to believe that the Crinoidea

diverged into two groups on what is a trifling point of meristic
variation comparable to the varying number of rows of plates in the
interradial areas of the older Echinoidea; and he is equally sceptical
as to the validity of Jaekel's division of the group into
Cladocrinoidea and Pentacrinoidea, leading to the view that
organs like pinnules represent totally different structures in different
groups. Wachsmuth and Springer adopt as bases of classification
the extent to which the arms and their branches are incorporated in
the disc, and they recognise three main divisions: Inadunata, in
which the arms are completely free from the calyx; Articulata, in
which the arms are partly incorporated but the tegmen remains
flexible; and finally Camerata, in which the arms and their first
branches are largely incorporated in the cup; the tegmen is
converted into a rigid dome and the ambulacral grooves on it
become closed, as does the mouth, by the meeting of overarching
folds; the grooves remaining, of course, open in the distal portions of
the arms (Fig. 273). This classification, founded as it is on
physiological factors, seems to the present author more satisfactory.
Speaking generally, the points in which fossil Crinoids may differ
from living genera are: (1) the total absence or irregular nature of the
branching in the arms, so that pinnules may be said to be absent; (2)
the closure of the ambulacral grooves and mouth already alluded to,
and (3) the adhesion of the arms in the same ray to produce net-like
structures (Crotalocrinus, Fig. 274), or a fan-shaped structure
(Petalocrinus); (4) the frequent presence of two rows of brachials in
one arm (biserial structure); (5) the development of an enormous
anal tube, so large that in extreme cases (Eucalyptocrinus) the arms
may be lodged in grooves of it.

CLASS II. THECOIDEA (EDRIOASTEROIDEA, Bather)

These remarkable Pelmatozoa are the most primitive known. They
have sac-like or sometimes cushion-shaped or even disc-shaped
bodies, covered with numerous irregular plates without any
symmetry in their arrangement. There is no stem, but when they are
fixed this is effected by an adhesion of the aboral pole. There are no
arms, but on the upper surface is to be seen the impression of five
ambulacral grooves radiating from a central mouth. These grooves
are bordered by covering plates, which in the earliest form
(Stromacystis) are seen to be slight modifications of the plates
covering the upper surface of the body, but in the later genera (Fig.
275, Thecocystis) become specialised. The anus is situated on the
side, as is also the madreporite. It has been suggested that
Eleutherozoa were derived from this group; that individuals were
occasionally overturned by the waves or currents, and in this way
compelled to use their podia for locomotion. When Eleutherozoa,
however, have a fixed stage in their development, they are fixed by
the oral, not the aboral, surface, and hence can have no close
affinity to Thecoidea. Thecoidea begin in the Middle Cambrian, but
according to Jaekel impressions in the Lower Cambrian, referred to
Medusae, may be casts of this group.

Fig. 275.—Thecocystis saeculus. × 6. a. Anus; m, mouth; p, madreporite(?)

(After Jaekel.)
 CLASS III. CARPOIDEA
Pelmatozoa with a well-developed stem; body bilaterally
compressed; only two rays apparently developed. These are
indicated only by grooves radiating from the mouth; but in some
cases slight horn-like outgrowths of some of the plates of the calyx
may support prolongations of the grooves.

This group, which, like the foregoing, commences in the Cambrian,

is perhaps more primitive than the Thecoidea in showing less
influence of the water-vascular system on the skeleton; but in the
presence of a differentiated stem and the development of only two
rays, it is more differentiated. The anus is on one of the flat sides,
covered with a flat plate acting as a valve. The members of this
group were formerly confounded with Cystoidea, from which they
differ in the absence of the characteristic pores. Trochocystis, the
genus figured, is devoid of any horn-like outgrowths of the calyx.

Fig. 276.—Trochocystis bohemicus, viewed from two sides. o, Mouth. (After

Jaekel.)

CLASS IV. CYSTOIDEA

Pelmatozoa with respiratory organs in the form of "diplopores" or
"pore-rhombs." In a great many cases there is a stalk, but in other
cases this is atrophied, and the animal is attached by the base of the
calyx. The radial canals run for a shorter or longer distance over the
calyx, but the plates of the calyx themselves are not modified for
them. Either they run in simple grooves, or they are protected by a
special series of plates lying above the plates of the calyx. The
terminal portions of the radial canals are in all cases free, supported
by unbranched arms consisting usually of a double row of ossicles.
These arms are termed "fingers."

It will be gathered from the description just given that the fingers and
the respiratory organs distinguish Cystoidea from the two foregoing
classes. Formerly this class was a lumber-room in which were
placed all primitive irregular Pelmatozoa. The labours of Jaekel[516]
have, however, dispelled the mist which enveloped this group, and in
his monograph all that can be extracted both from superficial
examination and dissection of these fossils is contained. It seems
possible to the present author that the class may eventually require
to be divided into two, corresponding to the two main divisions which
Jaekel recognises, viz. Dichoporita, with pectinated rhombs, and
Diploporita, with diplopores.

Fig. 277.—Echinosphaerites aurantium. A, from above; B, from the side; C,

neighbourhood of mouth, enlarged. amb, Ambulacral groove with side-
plates and covering plate; mad, madreporite. The short parallel lines across
the sutures are the "pore-rhombs." (After Zittel.)

The pore-rhombs of the Dichoporita (indicated in Fig. 277 by the

small parallel lines crossing the boundaries of the plates) were,
according to Jaekel, nothing but a series of folds of thin integument
projecting into the interior, the outer opening of which in most cases
adhered in the middle, leaving two pores connected by a groove.
The inner boundaries of the folds are sometimes preserved, but in
many cases they were entirely devoid of calcification, and so were
lost. The radial vessels either branched a great deal, giving rise to a
multitude of fingers, or, as in Echinosphaerites (Fig. 277), there were
a few long fingers supporting a reduced number of radial canals. In
some cases the calyx can be analysed into a regular series of cycles
of plates, consisting of basals, orals, and three intervening whorls,
thus including one more ring than the calyx of Crinoidea. Jaekel
regards this as a primitive arrangement, believing that the irregularity
seen in Echinosphaerites secondary. This is a doubtful hypothesis.

The diplopores of the Diploporita appear to consist of two canals

traversing the body-wall, opening close together into a common pit
externally, but diverging internally. Since in some cases, as in
Aristocystis (Fig. 278), this common pit is proved to have been
closed externally by a very delicate layer of calcification, it is
probable that the pores represent in other cases the points of origin
of finger-like gills similar to those of Asteroidea. Where they were
closed by calcification this was so thin and porous that the diffusion
through it sufficed for respiration. Jaekel regards the Diploporita as a
group derived from Dichoporita, but this seems to be extremely
doubtful.

Fig. 278.—Aristocystis. In the upper part of the calyx the heavy dots are
"diplopores," seen owing to removal of the superficial layer. (After Zittel.)

CLASS V. BLASTOIDEA
Pelmatozoa with respiratory organs in the form of longitudinal
calcified folds, termed "hydrospires," radiating from the mouth. Stem
well developed; calyx regular, consisting of a whorl of basals
surmounted by a whorl of forked radiais, in the clefts of which lay the
recumbent radial water-vascular vessels, supported each on a
special plate ("lancet plate"), and giving off two rows of branches
supported by short fingers (Fig. 279). Side-plates and covering
plates were also developed; five orals ("deltoids") completed the
calyx. The anus was at the side, just beneath one of the orals.

The hydrospires, which are the great characteristic of the class, are
seen in section in Fig. 279, B (hyd). They consist of a varying
number of parallel folds on each side of each "pseudambulacrum,"
as the lancet plate with its adhering side-plates and covering plates
has been termed. In the most primitive genus, Codaster, they appear
to have opened directly to the exterior, and to have been placed at
right angles to the lines of union of the radial and oral plates, just like
the grooves of a pectinated rhomb. In more modified forms, such as
Pentremites and Granatocrinus (Fig. 279), the outer openings were
overarched by the extension of the side-plates of the radial vessel,
and the whole group of folds has a common opening near the mouth;
indeed, in the highest form there is one common "spiracle" for the
two groups of folds in an interradius, which in one interradius is
confluent with the anus. The hydrospires, when they reach this form,
irresistibly recall the genital bursae of Ophiuroidea (Fig. 214, p. 490),
and very possibly served the same purpose.

Fig. 279.—Granatocrinus norwoodi. A, view of whole animal; B, section of

radius; C, an isolated finger. hyd, Hydrospire; l, lancet plate; pinn, finger;
p.p, covering plate; R and D both signify radial plate. (After Zittel.)

Reviewing the whole group of the Pelmatozoa, we see that in the

Cambrian they begin with the extremely primitive Thecoidea and
Carpoidea, together with some obscure forms which, combining a
stem with pentamerous symmetry in the calyx, are supposed to be
the forerunners of the Crinoidea. In the Lower Silurian or Ordovician
the two groups of the Cystoidea make their appearance, possibly
independently developed from either Carpoidea or primitive
Crinoidea, which in this period are present in unmistakable form. In
the Upper Silurian the Blastoidea appear, distinguishable from the
most regular Cystoidea only by their hydrospires. It seems practically
certain that they were developed from Cystoidea, and we follow
Jaekel in believing that they arose from Dichoporita. The Carpoidea
do not extend beyond the Ordovician, and by the end of the
Carboniferous period Cystoidea and Blastoidea die out, leaving only
the Crinoidea, which at that period were at their maximum
development. From the Carboniferous to the present day the
Crinoidea have continually decreased, leaving in recent seas, as
sole representatives of the Pelmatozoa, only the few forms
described at the beginning of this chapter.

CHAPTER XXI

ECHINODERMATA (CONTINUED): DEVELOPMENT AND PHYLOGENY

In Chapter XVI. it was stated that whilst a more or less perfectly

developed radial symmetry was one of the characteristic features of
the phylum Echinodermata when in the adult condition, yet in the
immature or larval condition the members of the group have a
strongly marked bilateral symmetry. In this feature larval
Echinodermata resemble the other Phyla of the animal kingdom
which have a well-developed coelom, such as Annelida, Mollusca,
Vertebrata, etc. Since, then, the peculiar radial symmetry is gradually
acquired during the growth of the Echinoderm, we may possibly
discover by a close scrutiny of the life-history what is the nature and
meaning of this departure from the ordinary type of structure among
coelomate animals.
There are two kinds of development met with amongst
Echinodermata, which may be roughly characterised as the
"embryonic" and the "larval" type respectively, although neither
description is exact. In developmental histories of the first type so
much reserve material is laid up in the egg in the form of food-yolk
that the young animal whilst in the bilateral stage requires little or no
food. In some cases, however, as in Amphiura squamata, the mother
pours out a nourishing exudation; but whether this is so or not, the
parent in nearly every case carries the young about with her until
they have reached the adult condition. In some Asteroidea, as for
instance in the Antarctic species Asterias spirabilis (Fig. 280), the
young become fixed to the everted lips of the mother; in Amphiura
squamata, and some other Ophiuroidea the eggs remain in the
genital bursae, which serve as nurseries; in some Spatangoidea, as
for instance in Hemiaster philippi (Figs. 250, 281), the eggs are
carried in some of the deeply grooved petaloid ambulacra; whilst in
Holothuroidea they may develop in the body-cavity (Phyllophorus
urna), or they may adhere to the back of the mother (Cucumaria
crocea, Fig. 259, p. 573), or they may be protected in special brood-
pouches either on the ventral side of the parent (Cucumaria
laevigata) or on the dorsal surface (Psolus ephippifer, Fig. 261).

The majority of these cases of embryonic development have been

recorded from Arctic or Antarctic waters; it appears as if conditions
there were not favourable to the larval type of development. In
Pelmatozoa the development of Antedon rosacea alone is known,
and that is of the embryonic type.

Fig. 280.—Oral view of Asterias spirabilis, slightly enlarged, showing embryos

attached to the everted lips. emb, Embryos. (After Perrier.)
So far, however, as their mode of propagation is known, it may
confidently be affirmed that the development of the majority of the
species of Eleutherozoa is of the second or larval type. In this type
there is little food-yolk in the egg, and the young animal or larva is
forced from a very early period of development to seek its own living,
and hence it is usually a considerable time (from a fortnight to two
months) before the adult form is attained. When the embryos of
different groups of Eleutherozoa are compared, there is no obvious
agreement in structure between them; but the larvae of the four
classes of Eleutherozoa exhibit with differences in detail a most
remarkable fundamental similarity in type, and we are accordingly
justified in regarding the larval development as primitive, and the
embryonic type as derived from it and differently modified in each
case.

Fig. 281.—Hemiaster philippi. Enlarged view of a single petal, showing the

embryos in situ. (From Wyville Thomson.) The whole animal is shown in
Fig. 250, p. 555.

In the typical larval development the eggs are fertilised after being
laid, and they then undergo segmentation into a number of equal, or
nearly equal, segments or "blastomeres." These arrange themselves
in the form of a hollow sphere or "blastula," the cavity of which is
called the "blastocoel" and afterwards becomes the primary body-
cavity of the larva. This cavity contains an albuminous fluid, at the
expense of which development appears to be carried on (Fig. 282,
B). The cells forming the blastula acquire cilia, and the embryo
begins to rotate within the egg-membrane, which it soon bursts, and,
rising to the surface of the sea, begins its larval life. The blastula is
therefore the first well-marked larval stage, and it is found in a more
or less recognisable form in life-histories of members of every large
group in the animal kingdom. Only in the case of Echinodermata and
of forms still lower in the scale, however, does it appear as a larval
stage. The free-swimming blastula stage is reached in from twelve to
twenty-four hours. Soon the spherical form of the blastula is lost; one
side becomes flattened and thickened, owing to a multiplication of
cells, so that they become taller and narrower in shape. Shortly
afterwards this thickened plate becomes buckled inwards,
encroaching on the cavity of the blastocoel. The larva has now
reached the second stage of its development; it has become a
"gastrula" (Fig. 282, C). The plate of thickened cells has become
converted into a tube called the "archenteron" (Fig. 282, C, arch),
which is the rudiment of both the alimentary canal and the coelom of
the adult. This tube communicates with the exterior, in virtue of its
mode of formation, by a single opening which is called the
"blastopore," which becomes the anus of the later larva and adult.
Whilst the gastrula stage is being acquired, the blastocoel or primary
body-cavity is invaded by wandering cells budded from the wall of
the archenteron (Fig. 282, A, B, C, mes). These cells, which are
called "mesenchyme," are the formative cells of the skeleton,
connective tissue, and wandering cells of the adult. When the larva
has a skeleton they are formed very early, arising in the young
blastula stage (Ophiuroidea) or in the stage immediately before the
formation of the archenteron (Echinoidea, Fig. 282, A, B) and
secreting the skeleton. When the larva is devoid of a skeleton
(Asteroidea and Holothuroidea), the mesenchyme usually does not
appear till the gastrula is fully formed.
 Fig. 282.—Echinus esculentus. A, optical section of living blastula. B, section of
preserved blastula. The network of strings in the interior is the result of the
coagulation of the albuminous fluid. C, section of gastrula. arch,
Archenteron; mes, mesenchyme cells, attached by protoplasmic strands to
the wall of the embryo. × 150.

The gastrula stage is reached in twenty to thirty-six hours. Then one

side of the larva becomes concave, and the cilia become restricted
to a thick band surrounding this area. In this way is formed the
rudiment of the longitudinal band of cilia, which is the organ of
locomotion throughout the larval life. At the apex of the archenteron
a thin-walled vesicle is formed, which soon becomes divided off from
the rest. This vesicle, which almost immediately divides into two
sacs, right and left, is the rudiment of the "coelom" or secondary
body-cavity of the larva; the remainder of the archenteron forms the
definitive gut, and becomes divided by constriction into an
oesophagus, a stomach, and an intestine, and at the same time bent
into a shallower or deeper V-shape, the concavity of which is
towards the concave side of the body. Within this area of the surface
a new funnel-shaped depression makes its appearance. This is the
"stomodaeum," the rudiment of the mouth of the larva, and it soon
joins the apex of the larval oesophagus; the conjoined tubes
henceforth bearing the name oesophagus since the ectodermal and
endodermal parts become indistinguishably fused. Along the sides
and floor of the oesophagus is formed a V-shaped ridge bearing
strong cilia; this is the "adoral band of cilia" which sweeps the food
(consisting of Diatoms, Infusoria, etc.) into the mouth. The larva is
now known as a Dipleurula and appears in four modifications, each
characteristic of a Class of Eleutherozoa. These differ from one
another principally in the following points:—(a) The folding of the
ciliated band; (b) the divisions of the coelomic sacs; (c) the
development and fate of the praeoral lobe (i.e. the part of the body in
front of the mouth); (d) the fate of the larval mouth. The types of
Dipleurula are as follows:—

Fig. 283.—Bipinnaria of Luidia. a, Anus; a.b, adoral ciliated band; a.c.o.b,

anterior median process; a.d.a, anterior dorsal process; a.v.a, prae-oral
process; m, mouth; p.c.o.b, median dorsal process; p.d.a, posterior dorsal
process; p.l.a, posterior lateral process; p.v.a, post-oral process. (After
Garstang.)

(1) The Bipinnaria, the larva of Asteroidea. In this type there is a very
long prae-oral lobe. The ciliated band runs along its edges, and is
produced into a backwardly directed loop on its under surface. This
loop soon becomes separated from the rest of the band as a distinct
prae-oral loop, the rest forming a post-oral loop. Both loops are
drawn out into short tag-like processes, in which we may distinguish
(following Mortensen's[517] notation) in the prae-oral loop an anterior
median process (Fig. 283, a.c.o.b), and a pair of prae-oral processes
(a.v.a). In the post-oral loop there is a median dorsal process
(p.c.o.b) and paired anterior dorsal (a.d.a), posterior dorsal (p.d.a),
posterior lateral (p.l.a), and post-oral (p.v.a) processes. At the apex
of the prae-oral lobe between prae-oral and post-oral ciliated rings
there is an ectodermic thickening, recalling the so-called apical plate
of Annelid larvae.
 Fig. 284.—A, Ophiopluteus of Ophiothrix fragilis. hy, Hydrocoel; l.p.c, left
posterior coelom; oes, oesophagus; r.p.c, right posterior coelom; st,
stomach. B, metamorphosis of Ophiopluteus of Ophiura sp. (After Johannes
Müller.)

(2) The Ophiopluteus, the larva of the Ophiuroidea. In this type the
prae-oral lobe remains small, and the primitive ciliated band is
undivided. The processes into which it is drawn out are very long,
and are supported by calcareous rods. Of these processes we may
distinguish prae-oral, postero-dorsal, postero-lateral, and post-oral.
The postero-lateral are always much longer than the rest, so that the
larva when swimming appears to the naked eye as a tiny V. In the
case of Ophiothrix fragilis (Fig. 284, A) the postero-lateral processes
are many times longer than the rest of the body. The Ophiopluteus
was the first Echinoderm larva to be recognised. It was discovered
by Johannes Müller,[518] who also discovered the other three types
of Dipleurula. He named this one Pluteus (easel), from a fancied
resemblance, when turned upside down, to a painter's easel. The
same name was bestowed on the next type, to which it presents a
superficial resemblance, and hence the distinguishing prefix "Ophio-"
was added to the original name by Mortensen.

(3) The Echinopluteus, the larva of the Echinoidea. This type is

strikingly like the preceding one in possessing a very small prae-oral
lobe and in having the processes of the ciliated ring supported by
calcareous rods, but a close inspection of these shows that they do
not exactly correspond to those of the Ophiopluteus. Thus we have
prae-oral, postero-dorsal, and post-oral processes (Fig. 285), but
usually no postero-lateral process, and when it does occur it remains
short. On the other hand, an antero-lateral process unrepresented in
the Ophiopluteus is constantly present, and in its later stage the
Echinopluteus develops, out of parts of the ciliated ring, horizontally-
placed crescentic ridges of cilia, which are termed ciliated epaulettes
(Fig. 285, a.cil.ep). There may even be, as in the larva of Echinus
esculentus, a second posterior set of these (Fig. 285, p.cil.ep). In the
older larva at the apex of the prae-oral lobe there is an ectodermic
thickening, at the base of which are developed nerve-cells and nerve
fibres constituting a larval brain (Fig. 285, ap).

Fig. 285.—Dorsal view of larva of Echinus esculentus. × 45. a.cil.ep, Anterior

ciliated epaulette; ap, apical plate or larval brain; ech, rudiment of Sea-
urchin; l.a.c, left anterior coelom; l.oes, larval oesophagus; l.p.c, r.p.c, as in
Fig. 284; p.cil.ep, posterior ciliated epaulette; r.a.c, right anterior coelom.

(4) The Auricularia, the larva of the Holothuroidea. This type

strikingly resembles the Bipinnaria in its external features. The prae-
oral lobe is well developed, and has on its under surface a
backwardly projecting loop of the ciliated band, which is not,
however, as in the Bipinnaria, separated from the rest of the band.
The processes of the band are much more faintly marked than in the
Bipinnaria, the anterior median, prae-oral, and median dorsal
processes being absent; but a pair of intermediate dorsal processes
are developed in the interspace between anterior and posterior
dorsal.
 Fig. 286.—Three views of metamorphosis of Auricularia of Synapta digitata. A,
fully grown Auricularia; B and C, stages in the metamorphosis. hy,
Hydrocoel; Int, intestine; l.p.c, left posterior coelom; O, fragments of ciliated
band which are invaginated into the stomodaeum, and coalesce to form a
ring round the mouth; oss, ossicle; pod, rudiment of feelers which here
spring directly from the hydrocoel; r.p.c, right posterior coelom; st, stomach;
w.v.r, rudiment of water-vascular radial canals; 1-5, corresponding pieces in
the three figures of the longitudinal ciliated band. (After Bury.) × 40.

In the Bipinnaria, Ophiopluteus, and Echinopluteus the coelomic

vesicle, after separation from the archenteron, divides into right and
left halves. The left then sends out a short dorsal process, which,
fusing with the ectoderm, acquires an opening to the exterior. This
opening is the primary madreporic pore, and the process of the left
coelomic sac, which is ciliated, is the pore-canal. In the Auricularia
the pore and pore-canal are formed before the division of the
coelom. In the Bipinnaria the right and left sacs subsequently fuse in
the front part of the prae-oral lobe. In the first three types of larva the
coelomic sac on each side then undergoes a segmentation into
anterior and posterior portions. At the hinder end of the anterior sac
on each side a swelling occurs. That on the left side is the
"hydrocoel," or rudiment of the water-vascular system (Fig. 287, A3,
l.hy); it quickly assumes a crescentic form, and gives off five blunt
outgrowths, which are the rudiments of the radial canals, and the
terminal tentacles. It remains in connexion with the anterior coelom
by a narrow neck, which later becomes the stone-canal. That on the
right side separates completely from the right anterior coelom; it
remains small, and forms the madreporic vesicle (Fig. 287, A3, r.hy)
of the adult. In the Ophiopluteus and in the larva of Asterina gibbosa
(v. infra) it occasionally takes on a form similar to that of the
hydrocoel; from which circumstance, as well as from the similarity in