Professional Documents
Culture Documents
Textbook Algorithms and Architectures For Parallel Processing 18Th International Conference Ica3Pp 2018 Guangzhou China November 15 17 2018 Proceedings Part Iii Jaideep Vaidya Ebook All Chapter PDF
Textbook Algorithms and Architectures For Parallel Processing 18Th International Conference Ica3Pp 2018 Guangzhou China November 15 17 2018 Proceedings Part Iii Jaideep Vaidya Ebook All Chapter PDF
https://textbookfull.com/product/algorithms-and-architectures-
for-parallel-processing-ica3pp-2018-international-workshops-
guangzhou-china-november-15-17-2018-proceedings-ting-hu/
https://textbookfull.com/product/computational-science-
iccs-2018-18th-international-conference-wuxi-china-
june-11-13-2018-proceedings-part-iii-yong-shi/
123
Lecture Notes in Computer Science 11336
Commenced Publication in 1973
Founding and Former Series Editors:
Gerhard Goos, Juris Hartmanis, and Jan van Leeuwen
Editorial Board
David Hutchison
Lancaster University, Lancaster, UK
Takeo Kanade
Carnegie Mellon University, Pittsburgh, PA, USA
Josef Kittler
University of Surrey, Guildford, UK
Jon M. Kleinberg
Cornell University, Ithaca, NY, USA
Friedemann Mattern
ETH Zurich, Zurich, Switzerland
John C. Mitchell
Stanford University, Stanford, CA, USA
Moni Naor
Weizmann Institute of Science, Rehovot, Israel
C. Pandu Rangan
Indian Institute of Technology Madras, Chennai, India
Bernhard Steffen
TU Dortmund University, Dortmund, Germany
Demetri Terzopoulos
University of California, Los Angeles, CA, USA
Doug Tygar
University of California, Berkeley, CA, USA
Gerhard Weikum
Max Planck Institute for Informatics, Saarbrücken, Germany
More information about this series at http://www.springer.com/series/7407
Jaideep Vaidya Jin Li (Eds.)
•
123
Editors
Jaideep Vaidya Jin Li
Rutgers University Guangzhou University
Newark, NJ, USA Guangzhou, China
This Springer imprint is published by the registered company Springer Nature Switzerland AG
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
Preface
of this conference. We would like to express our special appreciation to Prof. Yang
Xiang, Prof. Weijia Jia, Prof. Yi Pan, Prof. Laurence T. Yang, and Prof. Wanlei Zhou,
the Steering Committee members, for giving us the opportunity to host this prestigious
conference and for their guidance with the conference organization. We would like to
emphasize our gratitude to the general chairs, Prof. Albert Zomaya and Prof. Minyi
Guo, for their outstanding support in organizing the event. Thanks also to the publicity
chairs, Prof. Zheli Liu and Dr Weizhi Meng, for the great job in publicizing this event.
We would like to give our thanks to all the members of the Organizing Committee and
Program Committee for their efforts and support.
The ICA3PP 2018 program included two workshops, namely, the ICA3PP 2018
Workshop on Intelligent Algorithms for Large-Scale Complex Optimization Problems
and the ICA3PP 2018 Workshop on Security and Privacy in Data Processing. We
would like to express our sincere appreciation to the workshop chairs: Prof. Ting Hu,
Prof. Feng Wang, Prof. Hongwei Li and Prof. Qian Wang.
Last but not least, we would like to thank all the contributing authors and all
conference attendees, as well as the great team at Springer that assisted in producing the
conference proceedings, and the developers and maintainers of EasyChair.
General Chairs
Albert Zomaya University of Sydney, Australia
Minyi Guo Shanghai Jiao Tong University, China
Program Chairs
Jaideep Vaidya Rutgers University, USA
Jin Li Guangzhou University, China
Publication Chair
Yu Wang Guangzhou University, China
Publicity Chairs
Zheli Liu Nankai University, China
Weizhi Meng Technical University of Denmark, Denmark
Steering Committee
Yang Xiang (Chair) Swinburne University of Technology, Australia
Weijia Jia Shanghai Jiaotong University, China
Yi Pan Georgia State University, USA
Laurence T. Yang St. Francis Xavier University, Canada
Wanlei Zhou Deakin University, Australia
Program Committee
Pedro Alonso Universitat Politècnica de València, Spain
Daniel Andresen Kansas State University, USA
Cosimo Anglano Universitá del Piemonte Orientale, Italy
Danilo Ardagna Politecnico di Milano, Italy
Kapil Arya Northeastern University, USA
Marcos Assuncao Inria, France
Joonsang Baek University of Wollongong, Australia
Anirban Basu KDDI Research Inc., Japan
Ladjel Bellatreche LIAS/ENSMA, France
Jorge Bernal Bernabe University of Murcia, Spain
Thomas Boenisch High-Performance Computing Center Stuttgart,
Germany
VIII Organization
Facial Shape and Expression Transfer via Non-rigid Image Deformation . . . . 257
Huabing Zhou, Shiqiang Ren, Yong Zhou, Yuyu Kuang, Yanduo Zhang,
Wei Zhang, Tao Lu, Hanwen Chen, and Deng Chen
POEM: Pricing Longer for Edge Computing in the Device Cloud . . . . . . . . . 355
Qiankun Yu, Jigang Wu, and Long Chen
1 Introduction
With the explosive growth of smart phones, wearable devices and Internet of
Things (IoT), they are demanding for higher data rates and lower latency. Spec-
trum resource is one of the most valuable resources for wireless communication
devices. However, many spectrum resources have been allocated to licensed users.
On one hand, existing un-used spectrum resources have become scarce. On the
other hand, some used spectrum resources have not been fully utilized, such as
c Springer Nature Switzerland AG 2018
J. Vaidya and J. Li (Eds.): ICA3PP 2018, LNCS 11336, pp. 3–16, 2018.
https://doi.org/10.1007/978-3-030-05057-3_1
4 X. Zhang et al.
In this section, we not only focus on the system model, but also formulate the
problem to be studied. And we introduce the related economic properties that
auction scheme should be followed. The basic notations as shown in Table 1.
In this paper, we consider a cognitive network with multiple primary users and
multiple secondary users. Besides, in order to improve the channel transmission
rate, we take the relay node into account. In this scenario, as in [16], we assume
all nodes stay static in a given auction period. TAMSA scheme aims to maximize
the social welfare in a spectrum auction, which also encourages both P U s and
SU s to participate in. To maximize the utilization of spectrum resources, the
incentive mechanism should properly assign the matching between the spectrum
6 X. Zhang et al.
Notations Meaning
P Us Set of primary users
SU s Set of secondary users
Ri The ith relay node, where i ∈ [1, M ]
Si The ith group, where i ∈ [1, ni ]
sji The jth secondary user in the ith group, 1 ≤ i ≤ M, 1 ≤ j ≤ ni
dji (k) Demand of sji for kth Channel (P Uk ), 1 ≤ k ≤ M
bji (k) The bid of sji for kth Channel
vij (k) The valuation of sji for kth Channel
Ak Ask or reserve price of kth Channel
Siw Set of winning secondary users, 1 ≤ w ≤ ni
Riw Set of winning relay nodes
P Uiw Set of winning primary users
pji (k) The payment of sji for kth Channel
pc (k) The clearing price
Fi (k) Si (k) s payment for kth relay node
Pi (k) The ith relay node Ri (k) s payment for P Uk
Bi (k) The bid of the ith relay node Ri (k) for P Uk
uji The utility of sji
UP Uk The utility of P Uk
URk The utility of Rk
The system will determine the payment of winners. To achieve fairness, payments
of winners should be proportional to the workloads of the demands. The payment
of sji (k) is formulated as
where pc (k) is the clearing price. Let uji denote the utility of secondary user sji ,
for each sji ∈ Siw . Accordingly, the utility of sji is defined as
j
j vi (k) − pji (k), if sji ∈ Siw and pji (k) ≤ bji (k)
ui = (2)
0, otherwise.
What calls for special attention is that the payment of sji (k) should not be higher
than the budget bji (k), k ∈ [1, M ], sji ∈ Siw . The relay node Ri (k) calculates the
finance Fi (k) collected from SU s. Hence the utility of relay node Ri is
Fi (k) − Pi (k), if Ri (k) ∈ Riw
URi = (3)
0, otherwise.
Where Pi (k) is the payment of relay node for P U s. In order to encourage spec-
trum holders to share spectrum resources, each P Uk has a reserved price Ak .
The payment of relay nodes Pi (k) should be higher than the reserved price Ak ,
so the utility of P Uk is defined as
Pi (k) − Ak , if P Uk ∈ P Ukw and Ri ∈ Riw
UP Uk = (4)
0, otherwise.
In this auction, the spectrum owners P Uk s allocate spectrum resources to SU s.
The speed of channel transmission is increased by the relay nodes cooperatively.
In this stage, the ni secondary users are randomly divided into multiple groups.
The groups submit their bids or budgets to relay nodes separately. Relay nodes
will conduct the auction and decide the winning group members virtually. Then
relay nodes calculate the payment of each winner and determine the final win-
ners. It will allocate channels to SU s if it gets spectrum resources in tier II
auction.
We first introduce the algorithm to buy the spectrum by group and decide the
winners (GBDW), the details are as follows. Firstly, relay node Ri collects the
bid vector b1i , b2i , · · · , bni i , demand d1i , d2i , · · · , dni i and valuation vi1 , vi2 , · · · , vini
from SU s in Si as previous mentioned. We design an algorithm to calculate
the budget vector Fi (k) for P Uk . Then, relay nodes decide the winner in the
best performance ratio way and calculate the optimal unit price for each group.
The relay node Ri sells at most 1/2 time fraction to the Si for maximizing the
revenue. Inspired by the work in [16], we sort the vector of b/d in descending,
then we can get the optimal unit price for group Si , denoted as OP T (b/d),
bi
OP T (b/d) = max i , (5)
1≤i≤|b| di
where |b| denotes the length of the array, bi and di denote the ith budget and
demand separately. The detail of the algorithm is shown in Algorithm 1.
It should be noted that the clear price is extracted from the group to ensure
truthfulness. Relay nodes select the maximum integer m by OP T (b/d), and
then eliminate m SU s with smallest budget and lowest valuation. Fi (k) is the
gathered bid from those winning SU s, and the P Uk charges Ri (k) less than
Fi (k) for trading the kth channel.
In the example, we will show how Algorithm calculates the clearing price and
determines the winner. We assume that there are 5 SU s in group i, and their
budget and demands vector are as follows: b = {2, 3, 7, 6, 8}, d = {1, 2, 3, 2.5, 4},
so b/d = {2, 1.5, 2.33, 2.4, 2}, which can be obtained by Algorithm 1. We sort
b/d in descending and calculate OP T (b/d) to get the maximum m, hence we can
get m = 4 and the clearing price is pc = 8/4 = 2. Si participates in the auction
need to pay to the ith relay node is p1i = pc × d1i = 2 × 1 = 2. In the same way,
the payment of the other 4 secondary users can be calculated separately, which
is 4, 6, 5 and 8. Therefore, the winners in ith group are s1i , s3i , s4i and s5i , and the
amount collected by the ith relay node is 21.
10 X. Zhang et al.
4 Theoretical Analysis
In this section, we prove that TMASA satisfies the truthfulness, individual ratio-
nality, budget balance and computational efficiency.
Theorem 1. TAMSA is truthful in the network.
Proof. In the following, we focus on proving the dominant strategy for SU s. For
buyer, sji (k) ∈ Si , it will submit its true bid and demand, because it reflects its
true demand for spectrum resource.
TAMSA: Two-Stage Auction Mechanism for Spectrum Allocation 11
For sji ∈ Si , it can improve its utility by changing its valuation and budget
from the first branch of Eq. (2). Besides, inspired by [18], the clearing price pc is
randomly generated by the optimal price ratio.
For Ri ∈ Rw , it can obtain the maximum utility max(Bi (k) − Ak ) if it gets
the spectrum resource in this auction. If Ri ∈ / Rw , it will fail during this auction
and cannot get the spectrum resource, because (Bi (k)−Ak ) < 0. If relay node Ri
submits untruthful bid, the result will not change, when Bi (k) < Fi (k). When
Bi (k) > Fi (k), the utility of relay node URi (k) = Fi (k)−Pi = Fi (k)−Bi (k) ≤ 0,
if it submits untruthful bid.
Therefore, both relay nodes and SU s cannot improve their utility by sub-
mitting untruthful bids.
Proof. For SU s, the utility of sji (k) is calculated by vij (k) − pji (k) > 0, for ∀sji ∈
Siw , and we have proved the individual rationality of SU s. Then we prove relay
nodes are also individual rational. For relay node Ri , the minimum payment price
for relay node Ri (k) is Ak for ∀Ri ∈ Riw , Bik ≤ Fi (k) and P Uk ∈ P U w . Besides,
the utility of primary user UP Uk = Fi (k) − Pi ≥ Bi (k) − Ak > 0. Therefore,
both buyers and sellers are willing to participate in the auction. They can all
gain nonnegative utility, and TAMSA mechanism is of individual rationality and
budget balance.
5 Numerical Results
In this section, we evaluate the performance of TAMSA. In heterogeneous net-
work structure of we designed, this is the first incentive scheme proposed for
the specific demands of second users and there are no existing auction schemes
to compare with. Instead, We design the upper bound (Upper) and a random
algorithm (Random) for TAMSA to compare with. Meanwhile, we also simulate
algorithms TASG and TACC to compare with. The algorithm Upper uses the
bids of buyers as the payment to maximize the revenue. In TASG and TACC,
secondary users are divided into two sets randomly and selected the winning
set from other side. TASG is based on VCG mechanism, and TACC sorts the
reserve price Ak of primary users in ascending order and the budget Bi (k) of
relay nodes in descending order. The experiment tool is MATLAB, and the
results are averaged for 100 repetitions.
We consider a heterogeneous network shown in Fig. 1. We assume that the
number of P U s is M = 5, and there are 5 relay nodes to participate in this
auction, and the number of SU s ni varies from 20 to 120 with an increment
of 20. We assume that the valuation of secondary users vij (k) and budget bji (k)
are uniform distribution, and their ranges are denote as U (50, 150) and U (5, 10)
respectively. The reserve price Ak comply with U (10, 20) following [15–18].
We first investigate the running time of TAMSA, and the results are shown in
Figs. 2 and 3. From Fig. 2, we can see that the running time is no more than
0.35 s even if the amount of SU s becomes large, i.e., when there are 120 SU s. For
Fig. 3, we can see algorithm Random runs fastest, since the algorithm Random
selects winning secondary users Siw randomly.
0.4
Random
0.3 TAMSA
0.35
TASG
0.3 TACC
0.35 0.25
Running time(s)
0.3 0.25
0.25
Running Time(s)
0.2
0.2 0.2
0.15
0.15
0.1 0.15
0.05 0.1
0
30 0.1
0.05
120
20 100
Nu 80
mb 10 0
er 60 0.05
of Us 10 20 30 40 50
PU 40 of S
s 0 ber
20 Num Number of SUs
Fig. 2. Running time of TAMSA (a). Fig. 3. Running time of TAMSA (b).
TAMSA: Two-Stage Auction Mechanism for Spectrum Allocation 13
For TACC auction mechanism, the reserve price of primary user Ak is sorted
in ascending order, and the budget of relay nodes Bi (k) is in descending order to
guarantee the utility of P U s. Besides, TACC needs to match every primary user
and relay node, so algorithm runs the slowest. The running time of algorithm
TAMSA and algorithm TASG is not large, because they use maximum-weighted-
matching algorithm to complete matching between the winning P U s and relay
nodes.
Next, to validate Theorem 2 regarding individual rationality and budget
balance of TAMSA, we show the truthfulness in Fig. 4. In this auction, the
payment of relay nodes Pi (k) is not higher than the collected from SU s Fi (k),
and each winning primary user P Uiw receives a payment not less than its reserve
price Ak from the auctioneer.
From the experimental results in Fig. 4, we can see that the utility remains
nonnegative when relay nodes submit truthful bids. But when relay node submits
an untruthful bid, its utility rapidly reduce and will continue to be negative.
Figure 4 depicts the difference of utility. Relay nodes submit truthful bid, when
the bid of relay node is less than 50. When the bid is greater than 50, the
difference between truthful and untruthful bids is presented. The utility of relay
nodes and P U s are nonnegative, because the bid of relay node is less than
the collected from SU s, and the bid is greater than the reserve price of P U s,
Bi (k) ≤ Fi (k) and Bi (k) ≥ Ak , when relay nodes submit truthful bids. The
utility of relay node is negative, that is Bi (k) > Fi (k), if it submits an untruthful
bid. In summary, as seen in Fig. 4, the utility of relay nodes cannot be improved
by submitting untruthful bid.
80 9
Truthful Upper
unTruthful Random
60 8
TAMSA
TACC
7 TASG
40
Utility of relay nodes
6
Utility of PUs
20
5
0
4
−20
3
−40 2
−60 1
10 20 30 40 50 60 70 80 90 100 20 30 40 50 60 70 80 90 100
Bids of relay nodes Number of SUs
Figure 5 shows how the utility of primary users UP Uk varies with the number
of SU s. With the increasing number of SU s, the average utility of P U s calcu-
lated by the five algorithms is gradually increasing. On average, the proposed
algorithm TAMSA in this paper has improved 256% on the utility of P U s com-
pared with the algorithm Random. TASG is about 217% better than algorithm
14 X. Zhang et al.
Random, TACC achieves about 156% utility gains than the algorithm Random
on the UP Uk . TAMSA is further improved up by 28.33% and 78.65% over TASG
and TACC in terms of average utility of P U s, respectively. That’s because both
algorithm TAMSA and TASG apply the maximum weighted matching algorithm
to match P U s and relay nodes to ensure the maximum benefit. Besides, both
TAMSA and TASG use the auction mechanism based on VCG to ensure the
truthfulness of algorithm. The difference between TASG and TAMSA is that
TAMSA selects the winning set of SU s with the optimal cost performance, and
TASG selects the winning set with the subset of SU s’s bid by another subset. The
optimal cost performance can enhance the revenue of P U s. In TACC, although
the utility of P U s can be increase, it cannot guarantee the maximization of its
earnings.
50 4.5
Upper Upper
45 Random Random
TAMSA 4 TAMSA
40 TACC TACC
TASG TASG
35 3.5
Utility of Relays
30
Utility of SUs
3
25
2.5
20
15 2
10
1.5
5
0 1
20 30 40 50 60 70 80 90 100 20 30 40 50 60 70 80 90 100
Number of SUs Number of SUs
Fig. 6. Average utility of Relay nodes Fig. 7. Average utility of SUs with the
with the number of SUs. number of SUs.
Figure 6 depicts the average utility of relay nodes with the varying number of
SU s. We can see that TAMSA outperforms Random by about 10 times averagely,
TASG and TACC are about 7 times and 6.6 times better than Random algorithm
respectively. TAMSA is further improved up by 44.59% and 64.22% over TASG
and TACC in terms of average utility of relay nodes, respectively. The reason is
that both TAMSA and TASG use the VCG auction mechanism to calculate the
payment of relay nodes Pi (k). In Algorithm 2, we see that the payment of relay
node is effectively reduced on the premise of guaranteeing the primary user’s
revenue, so the utility of relay nodes is improved.
Figure 7 shows the relationship between the average utility of SU s and the
number of SU s. The average utility of SU s in TAMSA outperforms Random by
213%, TAMSA is able to improve the average utility of SU s in TASG up to 181%,
and TACC achieves about 115% utility gain than the Random algorithm on the
utility of SU s. TAMSA is improved up by 16.99% and 85.73% over TASG and
TACC in terms of average utility of SU s, respectively. That’s because TAMSA
selects the winning set Siw in optimal cost performance. The payment of SU s is
calculated according to their specific demands, so TAMSA effectively improves
Another random document with
no related content on Scribd:
The Project Gutenberg eBook of Six little
Bunkers at farmer Joel's
This ebook is for the use of anyone anywhere in the United States
and most other parts of the world at no cost and with almost no
restrictions whatsoever. You may copy it, give it away or re-use it
under the terms of the Project Gutenberg License included with this
ebook or online at www.gutenberg.org. If you are not located in the
United States, you will have to check the laws of the country where
you are located before using this eBook.
Language: English
BY
ILLUSTRATED BY
WALTER S. ROGERS
NEW YORK
GROSSET & DUNLAP
PUBLISHERS
Copyright, 1923, by
GROSSET & DUNLAP
“Margy, will you look out on the porch and see if she’s there?”
“Yes, Vi, I will. But you ought to say please to me, ’cause mother
says——”
“All right then. Please look and see if she’s there,” begged Vi,
otherwise Violet Bunker. There were six of the little Bunkers. The
other four will be out presently.
Margy, who had been looking at picture books with her year-older
sister in a room off the porch, kindly dropped her book and started
for the door.
“If she’s there bring her in—please.” Violet laughed a little as she
added the last word. She remembered what Margy had started to
say about politeness.
Violet was piling up the books, for she had just thought of
something new to play, when Margy came hurrying back into the
room.
“She isn’t there!” gasped the smaller Bunker girl.
“She isn’t?” Violet fairly gasped out the words, and you could
easily tell that she was very much excited. “Are you sure, Margy?”
“No, she isn’t there, Vi! Maybe a tramp has taken her!”
“Oh!” cried Violet, in such a loud voice that Mrs. Bunker, having
heard part of the talk, came quickly from the room where she had
been sewing.
“Who’s gone?” demanded the mother of the six little Bunkers.
“Don’t tell me Mun Bun is lost again!”
Mun Bun was the youngest of the six little Bunkers. His real name
was Munroe Ford Bunker, but that was entirely too long for the little
fellow, so he was called “Mun Bun.” It was a name he had made up
for himself.
“Where is Mun Bun? Is he lost again?” asked Mrs. Bunker, starting
to take off her apron to go in search of the “little tyke,” as she often
called him, for he certainly did get into mischief very many times.
“Mun Bun isn’t lost,” answered Violet, as she hurried out on the
porch with Margy. “He’s out in the yard with Laddie, digging a hole.”
“An’ he says he’s going to dig down to China,” added Margy.
“And I just put clean bloomers on him!” sighed Mrs. Bunker. “But
who is gone?” she asked again. “It can’t be Rose or Russ—they’re
too old to be taken by a tramp!”
There, now you have heard the names of all six of the little
Bunkers, though Russ, being nearly ten, I think, wouldn’t like to be
called “little.”
“No, it isn’t Russ or Rose,” said Margy. “I saw them going down
the street. Maybe they’re going to daddy’s office to ask him for some
money to buy candy.”
“Oh, they mustn’t do that!” exclaimed Mrs. Bunker. “This is the first
of the month and daddy is very busy. They shouldn’t have gone
there. Are you sure, Margy?”
“Oh, they didn’t zactly say they were going there,” announced
Margy. “But I thought maybe——”
“You mustn’t tell things you aren’t sure of,” said her mother. “But
who is——”
“Mother, why is daddy so busy the first of the month?” asked Vi,
forgetting for the moment all about what she had sent Margy to look
for. Violet Bunker was, as her father said, “a great girl for asking
questions.” Her mother knew this, and, fearing that Vi would get
started on a list of inquiries that would take some time to answer,
Mrs. Bunker said:
“Now don’t begin that, Vi, dear. I’ll answer just this one question,
but not any more. Your father is busy the first of the month more than
at other times because tenants pay their rents then, and he collects
the rents for a large number of people. That’s one thing a real estate
dealer, like your father, does. Now, don’t ask another question!” she
commanded, for she saw that Vi was getting ready, as Russ would
say, “to spring another.”
“I wasn’t going to ask a question,” said Vi, looking a little hurt in
her feelings. “I was going to say——”
“Wait until I find out what’s happened first,” broke in Mrs. Bunker.
“Who is missing? It can’t be any of you, for you’re all present or
accounted for, as they say in the army. Who is——”
“It’s Esmeralda!” exclaimed Violet. “I had her out on the porch
playing with Margy. Then we went in to look at the picture books, and
I forgot about Esmeralda and——”
“Russ says her name ought to be Measles ’cause she’s all
spotted,” put in Margy, with a shake of her dark, tousled hair. “But it’s
only spots of dirt.”
“Come on,” demanded Vi of Margy, taking her younger sister by
the hand. “We’ve got to find Esmeralda!”
“Oh, it’s your doll!” remarked Mrs. Bunker, with a sigh of relief. “I
thought one of you children was missing. I had quite a start. It’s only
your doll. That’s different.”
“Esmeralda is my child, even if she is only a doll,” and Vi marched
away with Margy, her head held up proudly.
“Oh, my dear, I didn’t mean that you shouldn’t want to find your
missing play child,” called Mrs. Bunker quickly, for she realized that a
little girl’s feelings might be hurt by a slighting remark about even a
dirty and spotted doll. “I only meant that I was glad none of you
children was missing. I’ll help you look for Esmeralda.”
“She isn’t out on the porch. I looked,” said Margy.
“We left her there, didn’t we?” asked Vi, for sometimes there was
so much going on at the Bunker house that to remember where one
of the many dolls or other playthings was left became a task.
“Yes, we left Esmeralda out on the porch,” agreed Margy. “But she
isn’t there now. I looked. She’s—she’s gone!”
Margy felt almost as sad over the loss as did Vi, though
Esmeralda, or “Measles,” as Russ called her, belonged particularly to
Violet.
“Do you s’pose a tramp would take my doll, Mother?” asked Violet,
for Mrs. Bunker was now walking toward the side porch with her two
little girls.
“No, my dear, I don’t believe so,” was the answer. “What would a
tramp want with a doll?”
This puzzled Vi for a moment, but she quickly had ready a reply.
“He—he might want to give her to his little girl,” Vi said.
“Tramps, as a rule, don’t have little girls,” remarked Mrs. Bunker.
“If they had they wouldn’t be tramps.”
This gave Vi a chance to ask another question. Eagerly she had it
ready.
“Why don’t tramps have little girls?” she inquired of her mother.
“Do they run away? I mean do the little girls run away?”
“No, that isn’t the reason,” and Mrs. Bunker tried not to smile at
Vi’s eagerness. “I’ll tell you about it some other time. But show me
where you left your doll,” she added, as they reached the shady side
porch. “Esmeralda certainly isn’t here,” for a look around showed no
doll in sight.
“Oh, where can she be?” gasped Vi, now on the verge of tears.
Margy, seeing how her sister was affected, was also getting ready to
weep, but just then a merry whistle was heard around the corner of
the house. It was the merry whistle of a happy boy.
“Here comes Russ!” exclaimed Violet, for she knew her oldest
brother’s habit of being tuneful. “He’ll help me look for Esmeralda.”
“Maybe he took her,” suggested Margy.
“No. If he did he wouldn’t be coming back whistling,” decided Vi.
Russ Bunker, next to his father the “man” of the family, swung
around the path at the side of the house. Following him was Rose,
his sister, a year younger, a pretty girl, with light, fluffy hair. And, very
often, Rose had a merry song on her lips. But as Russ was now
whistling Rose could not sing. She always said Russ whistled “out of
tune,” but Russ declared it was her singing that was off key.
“Oh, Russ!” exclaimed his mother, “you didn’t go to daddy’s office
and bother him to-day, did you, when it’s the first of the month? And
he is so busy——”
“No, Mother, I wasn’t at daddy’s office,” Russ answered. “Rose
and I just went to the store for some nails. I’m making a seesaw, and
——”
“Oh, can I be on it?” begged Margy. “I love to teeter-totter! Please,
Russ, can’t I——”
“I want a ride, too!” put in Vi.
“All right! All right!” agreed Russ, with a laugh. “You can all have
rides—Mun Bun and Laddie too—as soon as I get it made. But it’s a
lot of work and it’s got to be done right and——”
Russ paused. He could see that something was wrong, as he said
afterward. Russ was a quick thinker. Also he was always making
things about the house. These were mostly things with which to play
and have a good time, though once he built a bench for his mother.
The only trouble was that he didn’t make the legs strong enough,
and when Norah O’Grady, the cook, set a tub of water on the bench
the legs caved in and there was a “mess” in the kitchen.
“Has anything happened?” asked Russ, for he could see that his
mother and his two small sisters had come out on the porch with
some special idea in mind.
“Violet’s doll is gone,” explained Mrs. Bunker. “She left it on the
porch, and she feels sad over losing it. If you know anything about it,
Russ——”
“You mean that old Measles doll?” asked the oldest Bunker boy,
laughing.
“She hasn’t the measles at all—so there!” and Violet stamped her
foot on the porch.
“Well, she looks so—all spotted,” added Russ, with another laugh.
Then, as he saw that Violet was ready to cry and that Margy was
going to follow with tears, Russ added: “I guess I know where your
doll is. Henry Miller just told me——”
“Oh, did he take her?” cried Violet. “If he did I’ll never speak to him
again and——”
“Now, wait a minute!” advised Russ. “You girls always get so
excited! I didn’t say Henry took your doll. I just met him and he said
he saw a dog running out of our yard with something in his mouth.
Maybe it was the dog that took your doll, Violet.”
“Oh! Oh!” cried the little girl, and she was now sobbing in real
earnest.
“Oh, the dog will eat up Esmeralda!” and Margy added her tears to
those of Violet.
“I’ll go down the street and look for her,” quickly offered Russ. He
was a kind boy that way. Of course he didn’t care for dolls, and he
was anxious to start making the seesaw, nails for which he and Rose
had gone after. But Russ was willing to give up his own pleasure to
help his little sister.
“I’ll get your doll,” he said. “I guess that dog wouldn’t carry her far
after he found out she wasn’t a bone or something good to eat.”
“She—she—she’s a nice doll, anyhow, so there!” sobbed Violet.
“An’—an’ I—I want her!”
“I guess I can find her,” offered Russ. “Here, Rose, you hold the
nails.”
Russ started on a run toward the front gate. Mrs. Bunker and the
three girls followed. As yet Laddie and Mun Bun had not heard the
excitement over the missing doll, for they were still in the back yard,
“digging down to China.”
Russ reached the gate, looked down the road in the direction
Henry Miller had told him the dog had run with something in its
mouth, and then Russ cried:
“I see her! I see your doll, Vi! The dog dropped her in the street! I’ll
get her for you.”
Russ started on the run toward a small object lying in the dust of
the road. Before Russ could reach the doll a big automobile truck
swung around the corner and came straight for poor Esmeralda.
“Oh, she’ll be run over!” screamed Violet. “My child!”
But Russ had also seen the truck and, knowing there would be
little left of the doll if one of the heavy wheels went over her, he ran a
little faster and darted directly in front of the big lumbering,
thundering automobile.
“Russ! Russ! Be careful!” called his mother.
“Look out there, youngster!” yelled the man who was driving the
truck.
On came the heavy automobile, bearing down on Russ who was
now in the middle of the street, stooping over to pick up Esmeralda.
CHAPTER II
A LOAD OF FLOWERS