Sequencing has recently become a primary method used by life scientists to investigate biologically relevant problems related to genomics. As modern sequencers can only read very short fragments of the DNA strands, an algorithm is needed to assemble them into the original sequence. We propose a new algorithm based on this classical approach, but being able to accurately handle large data sets coming from next generation sequencing machines. We start with an unique way to construct the DNA overlap graph model by employing the power of alignment-free sequence comparison. The novelty of our solution lies in a special sorting technique that puts similar sequences close to each other without performing the sequence alignment. This phase is very fast and serves as preselection of similar pairs of sequences. Then two highly parallelizable steps proceed. Firstly, an ultra fast exact sequence comparison verifies previously selected candidates, resulting in very accurate results. The high performance computations employing both multiple CPUs and GPUs make the method very efficient even for large data sets. Secondly, having the DNA graph, the algorithm goes on to traverse it in a parallel way to obtain so called contigs and scaffolds, i.e. long fragments of reconstructed genome. Again the approach is novel, as resulting contigs are precisely cut in places where the repetitive fragments are detected.

Control Chart Patterns (CCPs) can be considered as time series. They are used in monitoring the control process, therefore an ability to recognize these patterns are essential in manufacturing as abnormalities can then be detected at early stage. Feeding CCPs directly to classifiers has proven unsatisfactory, especially in existence of noises. Therefore, there have been different kinds of preprocessing of CCPs to aid the classification. This research has two main objectives, first, is to study how the lengths of CCPs affect the performance of the classification. Second, the study attempts to determine the most suitable preprocessing techniques that have been applied to CCPs. Three preprocessing techniques are selected, these are Kalman filter, statistical features and symbolic representation known as Symbolic Aggregate ApproXimation (SAX). The Minimum Descriptive Length (MDL) algorithm for selecting SAX parameters is also investigated. Neural network is chosen as the tool to implement classifiers. The study concludes that longer patterns are more preferable than shorter ones. Statistical features is found as the best preprocessing techniques.

In the past years, cloud computing has seen an exponential growth. Finding the right set of cloud services among the numerous Cloud Service Providers (CSPs) that will best fulfll one customer's needs has thus become an arduous task. 

Cloud service brokers (CSBs) propose to assist customers in selecting the best services according to some criteria, e.g. cost and Quality of Service (QoS). The corresponding Cloud Brokering Optimisation (CBO) problem has given rise to novel models, including a variant of the Internet Shopping Optimization Problem (ISOP). As for the original problem, the NP-hardness of the CBO problem motivates the development novel efficient optimisation methods.

This work proposes to tackle the CBO problem with a Coevolutionary Genetic Algorithm (CGA). Contrary to standard Evolutionary Algorithms (EAs) that evolve one population of homogeneous individuals representing the global solution, coevolution considers several subpopulations representing subparts of the global solution that either compete or cooperate. Such a decomposition is proposed for the CBO problem and the corresponding CGA performance is experimented on a set of CBO benchmark instances.

Grégoire Danoy received his Industrial Engineer degree in Computer Science from the Luxembourg University of Applied Sciences (IST) in 2003. He obtained his Master in Web Intelligence in 2004 and his PhD in Computer Science in 2008 from the Ecole des Mines of Saint-Etienne, France. Since 2008, Dr. Danoy is Research Scientist in the Computer Science and Communications research unit (CSC) of the University of Luxembourg. His current research interests include nature inspired algorithms and multi-agent systems for tacklng telecommunications, mobile networks, bioinformatics, high performance and cloud computing problems. Dr Danoy has published more than 50 research articles in the eld and co-authored one book on evolutionary algorithms for mobile ad hoc networks (Wiley).

A Cloud Service Broker (CSB) is defined by the International Organization for standardization as a "cloud service partner between Cloud Service Customers (CSCs) and Cloud Service Providers (CSPs)". It selects the best services in terms of costs and quality to provide CSCs the best cloud computing solutions. Given the large number of providers and services, this task becomes hard and needs new Optimization solutions.

Some new models have been already proposed through the Internet Shopping Optimization Problem (ISHOP) which may be easily adapted to cloud brokering problems. Nevertheless, it is worthwhile to mention that those models only consider a single-level of decisions and suppose that CSBs and CSCs decide as a single entity. Bi-objective optimization may be used in such cases but game theory aspect would be lost. This is the reason why bi-level programming may be recommended. Indeed bi-level situations consist of two decision makers who control their own set of variables. The first decision maker referred to as the leader takes a decision which restricts the decision of the second decision maker referred to as the follower. In response to it, the follower will try to react optimally to the leader's decision. This modelling pattern may lead to collaboration or competition between them. Furthermore a bi-level strategy is more realistic since it does not overestimate the objective fitness when several decision makers may have an impact on each other. In the context of Cloud Brokering Optimization, CSBs want to guide the customer's decisions to maximize its profits while CSCs try to lower their costs. This last example can be reproduced using different objective functions (e.g. security aspects, green computing indicators).

In this paper we examine the reliability of subjective rating judgments along a single dimension, focusing on estimates of technical quality produced by integrity impairments and failures (non-accessibility, and non-retainability) associated with viewing video. There is often considerable variability, both within and between individuals, in subjective rating tasks. In the research reported here we consider different approaches to screening out unreliable participants. We review available alternatives, including a method developed by the ITU, a method based on screening outliers, a method based on strength of correlations with an assumed “natural” ordering of impairments, and a clustering technique that makes no assumptions about the data. We report on an experiment that assesses subjective quality of experience associated with impairments and failures of online video. We then assess the reliability of the results using a correlation method and a clustering method, both of which give similar results. Since the clustering method utilized here makes fewer assumptions about the data, it may be a useful supplement to existing techniques for assessing reliability of participants when making subjective evaluations of the technical quality of videos.

We study the scheduling problem on parallel identical machines with a common due date and the total late work criterion. The late work criterion estimates the quality of a solution based on the duration of late parts of jobs. Jobs appearing in the system have to be scheduled on machines, preferably before the given due date, in order to minimize their late parts.

In the offline case all jobs are known in advance, while in the online case they arrive in the system one by one. Late work scheduling finds many practical applications. Among others, the late work criterion can be used to optimize the process of shipping orders in the Internet Shopping Optimization Problem. To ship an order, a worker (modeled by a machine) has to collect and pack all ordered items. This task is represented by a job. Usually orders are shipped in batches (e.g., loaded on the same vehicle), for which shipping dates are defined. The shipping date is represented by a common due date, defined for a given group of orders (i.e., a set of jobs).

The goal is to minimize the size of orders which are not ready for shipping at the required time. The late work criterion has not been studied in the online mode so far. Thus, the analysis of the online problem was preceded by the analysis of the offline problem, whose complexity status has not been formally stated in the literature yet. We proved the binary NP-hardness of the offline case for two identical machines by showing the transformation from the partition problem and proposing the pseudopolynomial time dynamic programming algorithm. Then, we proposed an online algorithm for an arbitrary number of machines. We determined its competitive ratio (i.e. the upper bound of the distance between the optimal offline solution and any online solution). Moreover, we proved the optimality of this algorithm for two machine case by showing the equality of its competitive ratio and the lower bound of a competitive ratio of any online algorithm.

Internet Shopping Optimization Problem is a problem where a customer wants to buy a set of products online, choosing them from many available shops, paying costs of items but also necessary delivery costs. New version of this problem will be introduced, where within the budget constraints, the user wants to receive a maximal number of items or maximal combined perceived value of the items. This way, an incomplete order realization is allowed. The problem resembles or even is a generalization of some other well-known optimization problems like the Multiple Knapsack Problem (MKP) or the Maximum Coverage Problem (MCP). Mathematical formulation of the problem will be presented and computational complexity will be analyzed. Finally, an efficient algorithm for solving the problem will be described.