Qatar Foundation Annual Research Forum Volume 2012 Issue 1

Background and Objectives: There is high requirement of face and signature based multimodal biometric systems in various areas, such as banking, biometric systems and secured mobile phone operating systems. Few studies have been carried out in this area to enhance the performance of identification and authentication based on the fusion of those modalities. In multimodal biometric systems, the most common fusion approach is integration at the matching score level, but it is necessary to compare this strategy of fusion to the other strategies, like fusion at feature level. Our system combines these two biometric traits and provides better recognition performance compared with single biometric systems. Multimodal Authentication Systems: The first monomodal verification system is based on face verification using Gabor filters for feature extraction. The second system is based on online signature verification using Nalwa's method. The classification is released using the Cosine Mahalano-bis distance. Due to its efficiency, we used max-of-scores strategy to fuse face and online signature scores. The second proposed system is based on fusion at the feature level. Results and Conclusions: The performance of feature-level fusion and max-of-scores fusion techniques using face and online signature modalities were evaluated on ORL face database and the QU signature database. The lowest equal error rate is obtained by using a fusion strategy based on max-of-monomodal systems scores. Additionally, feature-level fusion based methods demonstrate a low equal error rate compared with the monomodal systems and have not been affected by the increase in the features vector dimension in term of time of verification; on the contrary, the fusion at the score level is clearly affected and it takes more in-time verification because it's necessary to get scores from each biometric trait before the fusion step.

Practical algorithms have been proposed to efficiently recompute the logical consequences of a Datalog program after a new fact has been asserted or retracted. This is essential in a dynamic setting where facts are frequently added and removed. Yet while assertion is logically well understood as incremental inference, the monotonic nature of traditional first-order logic is ill-suited to model retraction. As such, the traditional logical interpretation of Datalog offers at most an abstract specification of Datalog systems, but has tenuous relations to the algorithms that perform efficient assertions and retractions in practical implementations. This work proposes a logical interpretation of Datalog based on linear logic. It not only captures the meaning of Datalog updates, but also provides an operational model that underlies the dynamic changes of the set of inferable facts, all within the confines of logic. We do this specifically by explicitly representing the removal of facts and enriching our linear logic interpretation of Datalog inference rules with embedded retraction rules. These retraction rules are essentially linear implications designed to exercise the retraction of consequences when base facts are targeted for retraction. As such, we can map Datalog assertion and retraction onto the forward-chaining fragment of linear logic proof search. We formally prove the correctness of this interpretation with respect to its traditional counterpart. In the future, we intend to exploit our work here to develop a rich logic programming language that integrates Datalog style assertion and retraction with higher-order multiset rewritings.

Cloud computing has become a very popular platform to deploy data-intensive scientific applications, but this process faces its own set of challenges. Given the complexity of the application execution environment, routine tasks on the cloud such as monitoring, performance analysis, and debugging of applications become tedious and complex. These routine tasks often require close interaction and inspection of multiple layers in the cloud, which traditional performance monitoring tools fail to account for. In addition, many of these tools are designed for real-time analysis and only provide summaries of historical data. This makes it difficult for a user to trace the runtime performance of an application in the past. We present a new monitoring framework called All-Monitor Daemon (Almond). Almond keeps close tabs on cloud inventory by communicating with a cloud resource manager (such as VMware vCenter for a VMware private cloud). Almond then connects to each individual physical host in the inventory and retrieves performance metrics through the hypervisor. Examples of metrics include CPU, memory, disk and network usage. Almond is also designed to collect performance information from the Guest OS, allowing the retrieval of metrics from the application platform as well. Almond was designed from the ground up for enhanced scalability and performance. The framework uses a Time Series Database (TSD), and a decentralized monitoring architecture allows for fast performance queries while minimizing overhead on the infrastructure. Almond collects performance data from all the layers of the software stack, and collected data remains persistent for future analysis. As a result of our performance enhancements, our preliminary results indicate a 70% improvement in hypervisor query response time through these enhancements as compared to our previous monitoring solution, VOTUS. Almond is a work in progress, and will feature an intuitive web-based interface that allows system administrators and cloud users to view and analyze resources on the cloud. Once completed, Almond promises to be a highly scalable, fast performing and dynamic cloud resource monitor.

The MapReduce programming model is a widely accepted solution to address the rapid growth of the so-called big-data processing demands. Various MapReduce applications with a huge volume of input data can run on an elastic compute cloud composed of many computing instances. This elastic compute cloud is best represented by a virtual cluster, such as Amazon EC2. Performance prediction of MapReduce applications would help in understanding their scalability pattern. However, it is challenging due to the complex interaction of the MapReduce framework and the underlying highly-parameterized virtualized resources. Furthermore, MapReduce's high-dimension space of configuration paremeters which adds to the prediction complexity. We have evaluated a series of representative MapReduce applications on Amazon EC2, and identified how the cluster size affects the execution times. The scaling curve of all applications are studied to discover the scalability pattern. Our major findings are as follows: (1) The execution times of MapReduce applications follow a power-law distribution, (2) For map-intensive applications, the power-law scalability starts from a small cluster size, and (3) For reduce-intensive applications, the power-law scalability starts from a lager cluster size. We attempted to fit our scalability performance results using three regression methods: polynomial regression, exponential regression and power regression. By measuring the Root Squared Mean Error (RSME), the power regression performs best at performance prediction compared with the other methods evaluated. This was the case across all the benchmark applications studied. Our performance prediction methods will aid cloud users in choosing appropriate computing resources, both virtual and physical, from small-scale experimental test runs for cost saving.

In this paper, we consider a spectrum sharing network with two-way relaying and multiuser diversity. More specifically, one secondary transmitter with the best channel quality is selected and splits its partial power to relay its received signals to the primary users by using the amplify-and-forward relaying protocol. We derive a tight approximation for the resulting outage probability. Based on this formula, the performance of the spectral sharing region and the cell coverage are analyzed. Numerical results are given to verify our analysis and are discussed to illustrate the advantages of our newly proposed scheme.

Power Electronics converters are finding growing applications in industries and house hold devices. The growing automation in industrial applications needs highly complex power electronics systems to process the electric power. The controls of power electronic converters are still posing challenges due to their required precision. Several control strategies are developed and reported in the literature including Pulse Width Modulation (PWM), model predictive control, sliding mode control etc. The development of computational intelligence techniques are artificial intelligence, fuzzy logic, adaptive neuro-fuzzy inference system, genetic algorithms etc. The basic idea is to incorporate the human intelligence in the control system so it makes intelligent decisions. The paper presents an over-view of the application of computational intelligence techniques for control of power electronic converters. Computational Intelligence (CI) has been in focus for quite long time, and it is well known that CI techniques can help in solving complex multidimensional problems which are difficult to solve by conventional methods. Computational Intelligence technology is growing rapidly and its applications in various fields is being tested. Power electronics converters is one of the major application areas where this technology can play a vital and decisive role. Recent development of powerful digital signal processors and field programmable gate arrays is making implementation of Computational Intelligence technology economical with improvement of performance, compact and more competitive. Evidently, the future impact of this CI technology on power electronics converter control is very significant and utilizable.

Driven by the increasing and successful prevalence of MapReduce as an analytics engine on the cloud, this work characterizes the Map phase in Hadoop MapReduce to guide its configuration and improve overall performance. MapReduce is one of the most effective realizations of large-scale data-intensive cloud computing platforms. Hadoop is an open source implementation of MapReduce and is currently enjoying wide popularity. Hadoop has a high-dimensional space of configuration parameters (~200 parameters) that poses a burden on practitioners, like computation scientists, system researchers, and business analysts, to set for efficient and cost-effective execution. In this work we observe that MapReduce application performance is highly influenced by Map concurrency, defined in terms of two configurable parameters, the number of available map slots and the number of map tasks running over the slots. As Map concurrency is varied, we show that some inherent MapReduce characteristics allow systematic and well-informed prediction of MapReduce performance response (runtime increase or decrease). We propose Map Concurrency Characterization, MC2, a predictor for MapReduce performance response. MC2 allows for optimized configuration of the Map phase and, consequently, enhanced Hadoop performance. Current related schemes require mathematical modeling, simulation, dynamic instrumentation, static analysis of unmodified MapReduce application code, and/or actual performance measurements. In contrast, MC2 simply bases its decisions on MapReduce characteristics that are affected by Map concurrency. We implemented MC2 and conducted comprehensive experiments on a private cloud and on Amazon EC2 using Hadoop 0.20.2. Our results show that MC2 can correctly predict MapReduce performance response and provide up to 2.3X speedup in runtime for the tested benchmarks. This performance improvement allows MC2 to further serve in reducing cost in a cloud setting. We believe that MC2 offers a timely contribution to the data analytics domain on the cloud, especially as Hadoop usage continues to grow beyond companies like Google, Microsoft, Facebook and Yahoo!.

Background: According to the Qatar National Vision 2030, Qatar residents are encouraged to implement information and communication technology (ICT) initiatives in government, business, and education in pursuit of a knowledge-based society that embraces innovation, entrepreneurship, and excellence in education. This research project, which is funded by the Qatar National Research Fund (QNRF) under the National Priority Research Program (NPRP), is contributing to this vision by investigating the use of innovative training technology to train Qataris so that they are prepared for the 21st century workforce. Specifically, this research project investigates the use of mobile computer technology--such as mobile phones, tablet computers, and handheld computers--to train Qatar residents on workplace English so that they can become more effective when communicating in the workplace. This presentation will share the results of a preliminary study that was conducted. This project will be expanded using the "Framework for the Rational Analysis of Mobile Education" (FRAME) model (Figure 1) that describes the convergence of mobile technologies, human learning capacities, and social interaction. Objectives: The research evaluates the effectiveness of the mobile computer technology training and transferability to the Qatar workplace environment. Methods: A total of 27 trainees participated in this study. They were given a pre-test followed by the mobile learning training and then a post-test. Results: Overall, the learners' performance improved by 16 percent after completing the training with mobile technology. Ninety four percent of subjects said that the quality of the presentation on the mobile technology was either excellent, good, or fair. One hundred percent of subjects reported that the mobile technology helped them learn. Conclusion: The delivery of training using mobile computer technology was well received by learners. They liked the interactive and innovative nature of the training.

Qatar is now accumulating important expertise in biomedical data analytics. At QCRI, we are interested in providing for biomedical researchers based on their computational needs and in developing tools for data analytics in biomedical research. When computers extract patterns and classifiers from a body, they are used to predict new data that helps in defining a prior threat or disease. One non-invasive powerful technique for detecting and quantifying bio-markers linked to diseases (metabolites) is Nuclear Magnetic Resonance (NMR) spectroscopy. 1H NMR spectroscopy is commonly used in the metabolic profiling of biofluids. Metabolites in biofluids are in dynamic equilibrium with those in cells and tissues so their metabolic profile reflects changes in the state of an organism due to disease or environmental effects. The analysis of signals obtained from patients may be performed via methods which incorporate prior knowledge about the metabolites that contribute to the 1H NMR spectroscopic signals, recorded in a metabolite dataset. This paper presents a novel model and computationally automated approach that allows for the simulation of datasets of NMR spectra in order to test real data analysis techniques, hypotheses and experimental designs. Unlike others, this model generates NMR spectra of biofluids unlimited by the magnetic field or pH. It is simple to implement, requires small storage, and is easy to compute and compare. Moreover, it can treat metabolites with a relatively high number of 1H due to a special technique in programing based on physical properties. This model can open the door wide to a new technique of metabolite quantification and thus a better determination of metabolite concentrations which is the key of disease identification. The area of NMR expands rapidly and holds great promise in terms of the discovery of potential biomarkers of diseases, such as diabetes, an area of increasing concern in Qatar, and cancer, which is the third cause of death in Qatar.

Advances in wireless communications technologies (e.g., 3G, 4G and beyond) entail demands for higher data rates. A well-known popular solution to fulfill this requirement was to allocate additional bandwidth, which unfortunately is not anymore viable due to radio-frequency (RF) spectrum scarcity. Nonetheless, spectrum measurements around the globe have revealed the fact that the available spectrum is under-utilized. One of the most remarkable solutions to cope with the under-utilization of spectrum is the concept of cognitive radio (CR). In CR systems, the main implementation issues are spectrum sensing because of the uncertainties in propagation channel, hidden primary user (PU) problem, sensing duration and security issues. Hence, the accuracy and reliability of the spectrum sensing information can be suspicious and questionable inherently. There has been no study to date that investigates the impacts of absence of spectrum sensing information in CR networks. In this work, due to the imprecise and unreliable spectrum sensing information, we investigate the performance of an orthogonal frequency-division multiplexing (OFDM)-based (4G and beyond) CR spectrum sharing communication system that assumes random allocation and absence of the PU's channel occupation information, i.e., no spectrum sensing is employed to acquire information about the availability of unused subcarriers or the PU's activity. The results show that due to the lack of information of the PUs' activities, the SU randomly allocates the subcarriers of the primary network and collides with the PUs' subcarriers with a certain probability. The number of subcarrier collisions is found to be following hypergeometric distribution. The SU's capacity with subcarrier collisions is employed as a performance measure to investigate the proposed random allocation scheme for both general and Rayleigh channel fading models. To avoid the subcarrier collisions at the SUs due to the random allocation and to obtain the maximum sum rate for SUs based on the available subcarriers, an efficient centralized sequential algorithm is proposed and analyzed. The performance of such a communication set-up can provide various insights into the studies in the CR literature, and it these be utilized as a valid candidates for performance comparison benchmarks in CR spectrum sharing systems with the availability of spectrum sensing information.