Professional Documents
Culture Documents
Download textbook Computational Intelligence And Intelligent Systems 9Th International Symposium Isica 2017 Guangzhou China November 18 19 2017 Revised Selected Papers Part Ii Kangshun Li ebook all chapter pdf
Download textbook Computational Intelligence And Intelligent Systems 9Th International Symposium Isica 2017 Guangzhou China November 18 19 2017 Revised Selected Papers Part Ii Kangshun Li ebook all chapter pdf
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/
https://textbookfull.com/product/computational-intelligence-and-
intelligent-systems-kangshun-li/
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/
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/
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.)
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
•
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
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.
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.
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
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
A Novel Monitor Image De-hazing for Heavy Haze on the Freeway . . . . . . . 501
Chunyu Xu, Yufeng Wang, and Wenyong Dong
The Dynamic Relationship Between Bank Credit and Real Estate Price
in China. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 324
Xiaofan Wang and Li Zhou
Centralized Access Control Scheme Based on OAuth for Social Networks . . . 576
Yue Liu, Wei Gao, and Jingyun Liao
1 Introduction
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
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
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.
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
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)
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:
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 .
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.
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.
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
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.
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.)
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. 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.
CHAPTER XXI
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.
(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.