The style adopted in this document is close to the Harvard bibliography style. The differences are essentially in the in-text citation/referencing. The bibliography-Harvard style requires “et al.” to be used when there are more than three authors. The style used in this document adheres to the APA requirement, which requires “et al.” to be used when there are more than four authors. Both styles use “et al.” in subsequent citations after the first occurrence. In general, the aim is to provide accurate and consistent citations.
The purpose of an annotated bibliography is to help the researcher capture the essence of publishing.
works in an area of research interest. In addition, the document forms a basis for knowledge and eventual
critical review of the published literature on the research topic of interest. In the process of writing the
literature review, a natural output is pertinent to the research area. In essence, the researcher can
identify research questions and demonstrate familiarity with current solutions, along with the respective
strengths and weaknesses. It is logical to start the research by collecting and reading publications in the
general area of the topic being researched and producing an annotated bibliography as the initial output of
the research Endeavor. The following is a guideline on what the researcher should be looking for when
reading a given publication.
The annotations are predicated on finding answers to the following questions within the papers.
What problem is being solved? Definition of the problem.
What have others done about the problem(s)?
What solution is being proposed by the authors?
What result was obtained?
How does the solution/result compare/contrast with previous results?
What dataset was used in the experiments?
What conclusions and insights were offered?
What further work is proposed?
What is the relevance of the work to your current research?
Example usages of in-text citations
In this section, we provide examples of how in-text citations are used in several cases. In reading
In these examples, it is important to note the following:
The format of the first appearance of a citation is different from subsequent appearances.
Format of citation in “textual” mode is different from “parenthetical” mode.
Format of citation when there are single, two, three, four, five, and more authors are also different.
in the modes mentioned above. Bibliographic entries are also important to note, especially
when there are more than five authors.
Clarke (2011) said “ we can address fundamental philosophical questions about the nature of reality and
the mind. So, simply, science fiction can be described as the only genuine consciousness-expanding
the drug” (Clarke,2011) Notice the first citation is “textual” while the second is “parenthetical”.
Wu, J., Yu, Y., Huang, C. and Yu, K. (2015), Deep multiple instances learning for image clas- section and Kauto-annotation, in ‘Proceedings, IEEE Conference on Computer Vision and Pattern Recognition
l (CVPR)’, pp. 3460–3469.
Moreno, I. S., Garraghan, P., Townend, P., and Xu, J. (2014), ‘Analysis, modeling and simulate- the notion of
workload patterns in a large-scale utility cloud’, this research paper has three authors which are
listed above. The research paper Wu, J., Yu, Y., Huang, C. and Yu, K. (2015), Deep multiple
instance learning for image class- section and auto-annotation, in ‘Proceedings, IEEE Conference on
Computer Vision and Pattern Recognition (CVPR)’ also has three authors.
Dunbar, R. I. M., Arnab oldi, V., Conti, M. and Passarella, A. (2015) ‘The structure of online social networks
mirrors those in the offline world?’, This next citation of the paper (Awad et al., 2018) illustrates how
The subsequent citations should be set out when we have four authors.
Five and more authors
Zhang, X., Pham, D.-S., Venkatesh, S., Liu, W. and Phung, D. (2015), ‘Mixed-norm sparse representation
for multi-view face recognition. This research paper has five authors listed above and these could be
more. The research paper Zhang, X., Pham, D.-S., Venkatesh, S., Liu, W. and Phung, D. (2015),
‘Mixed-norm sparse representation for multi-view face recognition also has five authors.
Dunbar, R. I. M., Arnab oldi, V., Conti, M. and Passarella, A. (2015), ‘The structure of online social
networks mirror those in the offline world’, Social Networks 43, 39–47.
We use data on frequencies of bi-directional posts to define edges (or relationships) in two-Face
book datasets and a Twitter dataset and use these to create ego-centric social networks. We explore the
internal structure of these networks to determine whether they have the same kind of layered structure as
has been found in offline face-to-face networks (which have a distinctively scaled structure with
successively inclusive layers at 5, 15, 50, and 150 alters). The two Facebook datasets are best described by
a four-layer structure and the Twitter dataset by a five-layer structure. The absolute sizes of these layers
and the mean frequencies of contact with alter within each layer match very closely to the observed values.
from offline networks. In addition, all three datasets reveal the existence of an innermost network layer at
∼1.5 liters. Our analyses thus confirm the existence of the layered structure of ego-centric social networks.
with a very much larger sample (in total, >185,000 egos) than those previously used to describe them, as
well as identifying the existence of an additional network layer whose existence was only hypothesized in
offline social networks. In addition, our analyses indicate that online communities have very similar
structural characteristics to offline face-to-face networks.
Moreno, I. S., Garraghan, P., Townend, P. and Xu, J. (2014), ‘Analysis, j modeling and simulation of workload patterns in a large-scale utility k cloud’, IEEE Transactions on Cloud
Computing 2(2), 208–221.
Understanding the characteristics and patterns of workloads within a cloud computing environment
is critical in order to improve resource management and operational conditions while the Quality of Service
(QoS) guarantees are maintained. Simulation models based on realistic parameters are also urgently needed.
for investigating the impact of these workload characteristics on new system designs and operation policies.
Unfortunately, there is a lack of analyses to support the development of workload models that capture the
inherent diversity of users and tasks, largely due to the limited availability of Cloud trace logs as well as
the complexity of analyzing such systems. In this paper, we present a comprehensive analysis of the
workload characteristics derived from a production Cloud data center that features over 900 users
submitting approximately 25 million tasks over a time period of a month. Our analysis focuses on exposing
and quantifying the diversity of behavioral patterns for users and tasks, as well as identifying model
parameters and their values for the simulation of the workload created by such components. Our derived
model is implemented by extending the capabilities of the CloudSim framework and is further validated.
through empirical comparison and statistical hypothesis tests. We illustrate several examples of this work’s
practical applicability in the domain of resource management and energy efficiency.
Several works have investigated workload modeling using production usage data. Many such
research works use low-level application usage parameters, such as page accesses [11,18,22] or resource-level metrics [2,4]. However, in a highly complicated software system such as Enterprise Resource
Planning (ERP) software, using low-level concepts such as page access is too course-grained, as, for instance,
some pages might incur complex operations such as salary and pension calculations. Moreno et
al  study Google’s cloud trace logs to identify patterns in user requests. The authors defined cloud
workload in terms of “users” and “tasks”, where the user is a combination of submission rate, CPU, and memory
requested while the task is a combination of session length, average CPU, and memory utilization. We define
a high-level abstraction for clustering, which makes it applicable to other enterprise applications.s. While in
[15,18] the authors do consider a somewhat higher level of abstraction; they are still tied to the application
itself, rather than to the underlying business model. In our previous work , we surveyed a number of
workload generation methods and performance testing methods available and we describe the workload
and performance testing used at the case company in more detail
Wu, J., Yu, Y., Huang, C., and Yu, K. (2015). Deep multiple instances learning for image class- section and
auto-annotation, in ‘Proceedings, IEEE Conference on Computer Vision and Pattern Recognition
l (CVPR)’, pp. 3460–3469.
The recent development in learning deep representations has demonstrated its wide applications in
traditional vision tasks like classification and detection. However, there has been little investigation on how
we could build up a deep learning framework in a weakly supervised setting. In this paper, we attempt to
model deep learning in a weakly supervised learning (multiple instance learning) framework. In our setting,
each image follows a dual multi-instance assumption, where its object proposals and possible text
annotations can be regarded as two instance sets. We thus design effective systems to exploit the MIL.
property with deep learning strategies from the two ends; we also try to jointly learn the relationship
between object and annotation proposals. We conduct extensive experiments and prove that our weakly
Supervised deep learning framework not only achieves convincing performance in vision tasks, including
classification and image annotation but also extracts reasonable region-keyword pairs with little
supervision, on both widely used benchmarks like PASCAL VOC and MIT Indoor Scene 67; and also a
dataset for image-and patch-level annotations.
The problem of securely and efficiently searching cloud-stored encrypted documents from a mobile
device is addressed in this paper. Often, mobile device users outsource the storage of their data (photos,
email, documents, etc.) to a cloud service. A means of ensuring privacy and confidentiality is to encrypt
the data. This is especially so when the cloud service is semi-trusted. The problem addressed in this paper
is significant because of the need to search the cloud storage for required data without revealing what is
being searched or the keywords used for the search. Searchable encryption, which allows exact keywords
match and Boolean search, has previously been used for this task. The lack of inexact (fuzzy) match and
The ranking of burned search results limits the efficiency and usability of this scheme. The authors propose a
scheme that supports both fuzzy search and ranking. A fuzzy search based on wildcards has been previously
proposed, but it requires knowledge of the location of errors. Special-purpose error-tolerant encryption
scheme. also exists for iris data. Furthermore, the authors have previously devised a fuzzy search scheme
based on locality-sensitive hashing. Other proposed methods have memory burdens that are not suitable for
mobile devices. Ranking and fuzzy search have not previously been combined automatically in the setting.
considered in this paper. The proposed scheme uses an amplified locality-sensitive hashing method.
Specifically, the keyword encoding is performed by a piece-wise linear chaotic map and a relevance score.
along with a posting, lists are computed. These are protected by an order-preserving encryption scheme. The
posting list enables the ranking of returned files. The authors performed extensive tests to select an appropriate
locality-sensitive hashing method and the effect of its parameters on search time were presented. Retrieval
performance was evaluated using precision and recall metrics on the Enron and RFC datasets. The results
are indicative of the effectiveness of the proposed scheme. Security analysis in the semi-trusted framework
was also conducted. The proposed encrypted search scheme does not support conjunctive or disjunctive
Zhang, X., Pham, D.-S., Venkatesh, S., Liu, W. and Phung, D. (2015), ‘Mixed-norm sparse K
representation for multi-view face recognition, Pattern Recognition 48(9), 2935–2946
Cloud computing environments can be characterized by the behavior of users, task execution length, and
resource utilization patterns. Understanding the dynamics of these characteristics is useful in optimizing
the performance of the cloud data center and providing an agreed quality of service to users. Trace logs
capture the dynamics of the data center with respect to user behavior, task execution, and resource
utilization. By analyzing the trace logs, important parameters required for modeling and simulating the
environment, and optimizing its operation can be derived. This paper presents an in-depth
parameterization of the cloud computing environment model based on the trace logs of a large-scale
production data center. It further presents a validated simulation of the model that incorporates parameters
of users and tasks. Previous works have neither used large-scale trace logs nor considered task and user
Most of the previous works only provided coarse-grained analysis of the trace logs. The authors used
the Google trace logs (version 2) which contain over 25 million tasks submitted by 930 users during a
period of one month. In their analysis, three parameters were defined for user behavior: submission.
rate, α; CPU requested, β; memory required, φ. The tasks were analogously parameterized by: length, χ;
average CPU usage, γ; memory usage, π.
These parameters were used to derive clusters for users. and tasks; thus identifying the diversity of users
and tasks. Each of the clusters was fitted to a known probability distribution whose parameters were derived.
from the relevant trace logs. The model thus created formed the basis of validated simulations, using
CloudSIM, of the cloud environment.
Evaluation of the simulation model was performed using empirical graphical comparisons and
statistical hypothesis testing. Some of the future work indicated by the authors includes: studying the
relationship between users and jobs submitted; energy consumption and reliability; and collaborating with the
developers of Cloud SIM to incorporate the proposed model. A possible drawback of the methodology
adopted in the analysis is the use of known probability distributions. Perhaps a Gaussian mixture model
could have provided a more accurate model of the diverse user and task profiles. Face recognition with
multiple views is a challenging research problem. Most of the existing works have focused on extracting
shared information among multiple views to improve recognition. However, when the pose variation is too
large or missing, ‘shared information may not be properly extracted, leading to poor recognition results.
In this paper, we propose a novel method for face recognition with multiple view images to overcome the
large pose variation and missing pose issue.
By introducing a novel mixed norm, the proposed method automatically selects candidates from
the gallery to best represent a group of highly correlated face images in a query set to improve the classification.
accuracy. This mixed norm combines the advantages of both sparse representation-based classification
(SRC) and joint sparse representation-based classification (JSRC). The ℓ1-norm from SRC and
ℓ2,1-norm from JSRC are introduced to achieve this goal. Due to this property, the proposed method
decreases the influence when a face image is unseen and has large pose variations in the recognition process.
And when some face images with a certain degree of unseen pose variation appear, this mixed norm will
find an optimal representation for these query images based on the shared information induced from
multiple views. Moreover, we also address an open problem in robust sparse representation and
classification, which uses ℓ1-norm on the loss function to achieve a robust solution. To solve this
formulation, we derive a simple, yet provably convergent algorithm based on the powerful alternative
directions method of multipliers (ADMM) framework. We provide extensive comparisons which
demonstrate that our method outperforms other state-of-the-art algorithms on CMU-PIE, Yale B, and
Multi-PIE databases for multi-view face recognition.