Güney, I.A. | Küçük, G. | Özcan, E.

Conference Object | 2014 | Lecture Notes in Electrical Engineering264 LNEE , pp.97 - 106

In Simultaneous Multi-Threaded (SMT) processor datapaths, there are many datapath resources that are shared by multiple threads. Currently, there are a few heuristics that distribute these resources among threads for better performance. A selection hyper-heuristic is a search method which mixes a fixed set of heuristics to exploit their strengths while solving a given problem. In this study, we propose learning selection hyper-heuristics for predicting, choosing and running the best performing heuristic. Our initial test results show that hyper-heuristics may improve the performance of the studied workloads by around 2%, on the aver . . .age. The peak performance improvement is observed to be 41% over the best performing heuristic, and more than 15% over all heuristics that are studied. Our best hyper-heuristic performs better than the state-of-the art heuristics on almost 60% of the simulated workloads. © 2013 Springer International Publishing Daha fazlası Daha az

Guney, I.A. | Yildiz, A. | Bayindir, I.U. | Serdaroglu, K.C. | Bayik, U. | Kucuk, G.

Conference Object | 2015 | Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)9137 LNCS , pp.409 - 421

Inmulti- andmany-core processors, a shared Last Level Cache (LLC) is utilized to alleviate the performance problems resulting from long latency memory instructions. However, an unmanaged LLC may become quite useless when the running threads have conflicting interests. In one extreme, a thread can make benefit from every portion of the cache whereas, in the other end, another thread may just want to thrash the whole LLC. Recently, a variety of way-partitioning mechanisms are introduced to improve cache performance. Today, almost all of the studies utilize the Utility-based Cache Partitioning (UCP) algorithm as their allocation policy . . .. However, the UCP look-ahead algorithm, although it provides a better utility measure than its greedy counterpart, requires a very complex hardware circuitry and dissipates a considerable amount of energy at the end of each decision period. In this study, we propose an offline supervised machine learning algorithm that replaces the UCP lookahead circuitry with a circuitry requiring almost negligible hardware and energy cost.Depending on the cache and processor configuration, our thorough analysis and simulation results show that the proposed mechanism reduces up to 5% of the overall transistor count and 5% of the overall processor energy without introducing any performance penalty. © Springer International Publishing Switzerland 2015 Daha fazlası Daha az

Kucuk, G. | Uslu, G. | Yesil, C.

Conference Object | 2014 | Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)8488 LNCS , pp.187 - 198

In a Simultaneous Multi-Threaded (SMT) processor environment, threads share datapath resources, and resource allocation policy directly affects the throughput metric. As a way of explicit resource management, resource requirements of threads are estimated based on several runtime statistics, such as cache miss counts, Issue Queue usage and efficiency metrics. Controlling processor resources indirectly by means of a fetch policy is also targeted in many recent studies. A successful technique, Speculative Instruction Window Weighting (SIWW), which speculates the weights of instructions in Issue Queue to indirectly manage SMT resource . . .usage, is recently proposed. SIWW promises better peformance results compared to the well-accepted ICOUNT fetch policy. In this study, we propose an alternative fetch policy that implements SIWW-like logic using a history-based prediction mechanism, History-based Predictive Instruction Window Weighting (HPIWW), avoiding any types of speculation hardware and its inherent complexity. As a result, we show that HPIWW outperforms SIWW by 3% on the average across all simulated workloads, and dissipates 2.5 times less power than its rival. © Springer International Publishing Switzerland 2014 Daha fazlası Daha az

Helvaci, S. | Senova, A. | Kar, G. | Gören, S.

Conference Object | 2018 | Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)10995 LNCS , pp.193 - 204

This work proposes a gamification approach to measure the driving behavior using the in-vehicle data and score drivers. Existing work largely focus on one functionality: either displaying vehicular info or scoring the driver. And some other work just provides navigation or Point of Interest (POI). In our work, we combine these features with minimal distraction for the driver. With this goal, we consider a system that interfaces to the vehicle bus and find the errors of the driver during the drive using multiple criteria. Furthermore, by providing achievements and leader boards, the driver is motivated to have a good score while driv . . .ing. To facilitate this analysis and to evaluate the system, we recorded two trips in real traffic. The results show that we achieve more than 95% accuracy between real-world scenario and the simulation. We also present POI feature that finds nearest preferred locations which are restaurants, hospitals, gas stations, pharmacies, car repair shops. © Springer International Publishing AG, part of Springer Nature 2018 Daha fazlası Daha az

Özcan, E.

Conference Object | 2006 | Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)3867 LNCS , pp.85 - 104

This paper presents an empirical study on memetic algorithms in two parts. In the first part, the details of the memetic algorithm experiments with a set of well known benchmark functions are described. In the second part, a heuristic template is introduced for solving timetabling problems. Two adaptive heuristics that utilize a set of constraint-based hill climbers in a co-operative manner are designed based on this template. A hyper-heuristic is a mechanism used for managing a set of low-level heuristics. At each step, an appropriate heuristic is chosen and applied to a candidate solution. Both adaptive heuristics can be considere . . .d as hyper-heuristics. Memetic algorithms employing each hyper-heuristic separately as a single hill climber are experimented on a set of randomly generated nurse rostering problem instances. Moreover, the standard genetic algorithm and two self-generating multimeme memetic algorithms are compared to the proposed memetic algorithms and a previous study. © Springer-Verlag Berlin Heidelberg 2007 Daha fazlası Daha az

Özean, E. | Bilgin, B. | Korkmaz, E.E.

Conference Object | 2006 | Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)4193 LNCS , pp.202 - 211

Hyperheuristics are single candidate solution based and simple to maintain mechanisms used in optimization. At each iteration, as a higher level of abstraction, a hyperheuristic chooses and applies one of the heuristics to a candidate solution. In this study, the performance contribution of hill climbing operators along with the mutational heuristics are analyzed in depth in four different hyperheuristic frameworks. Four different hill climbing operators and three mutational operators are used during the experiments. Various subsets of the heuristics are evaluated on fourteen well-known benchmark functions. © Springer-Verlag Berlin . . .Heidelberg 2006 Daha fazlası Daha az

Canbay, F. | Levent, V.E. | Serbes, G. | Ugurdag, H.F. | Goren, S. | Aydin, N.

Conference Object | 2016 | IFMBE Proceedings57 , pp.114 - 118

In medical applications, biomedical acquisition systems (BASs) are frequently used in order to diagnose and monitor critical conditions such as stroke, epilepsy, Alzheimer disease, arrhythmias and etc. Biomedical signals (BSs), which produce valuable information about the condition of various physiological subsystems in our body, can be obtained by using multi-channel BASs. Due to the time-varying behavior of physiological sub-systems, most of the BSs are expected to have non-stationary character. In order to derive desired clinical information from these non-stationary BSs, an appropriate analysis method which exhibits adjustable t . . .ime-frequency resolution is needed. The wavelet transform (WT), in which the time-frequency resolution can be adjusted according to the different parts of the signal, are widely used in the analysis of BSs. The discrete wavelet transform (DWT) is a fast and discretized implementation of classical WT and was employed as a feature extractor and de-noising operator for BSs in literature. However, due to the aliasing, lack of directionality and being shift-variance disadvantages, the DWT exhibits limited performance. A modified version of the DWT, which is named as Dual Tree Complex Wavelet Transform (DTCWT), is employed in the analysis of BSs and improved results are obtained. Therefore, in this study, considering the improvements in embedded system technology and the needs for wavelet based multi-channel real-time feature-extraction/de-noising operations in portable medical devices, the DTCWT is implemented as a multi-channel system-on-chip by using field programmable gate arrays. In proposed hardware architecture, for N input-channels, the DTCWT is implemented by using only one adder and one multiplier. The area efficiency and speed limits of proposed system are presented comparing with our previous approaches. © Springer International Publishing Switzerland 2016 Daha fazlası Daha az

Perente, O.K. | Serif, T.

Conference Object | 2018 | Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)10995 LNCS , pp.142 - 153

With the advancement of technology and telecommunication services, data consumption rates are increasing ever since. People for long have started using applications with the help of contextual information to improve their user experience. Thus, providing a cross-platform location service to further enrich such applications has become a necessity. For this purpose, numerous client-based indoor location systems on mobile devices are developed to perform this task. Nevertheless, most of the time these systems suffer from elimination of features from operating systems for security purposes. Indeed, with the current security trends, to e . . .nsure the privacy of mobile users, mobile operating system designers are progressively eliminating certain low-level features such as reading RSSI and introducing randomized MAC addresses. Thus, in this study, the authors propose, design and implement a server-based indoor positioning system to eliminate platform dependency and to provide the location detection in wide range of devices. The designed server-based system is scalable and platform independent; hence can run on virtually any family of smart device. Furthermore, the evaluation findings indicate that the proposed system performs in acceptable accuracy to client-based systems compared to more complex and costly implementations. © Springer International Publishing AG, part of Springer Nature 2018 Daha fazlası Daha az

Kucuk, G. | Basaran, C.

Conference Object | 2006 | Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)4148 LNCS , pp.256 - 266

Wireless sensor networks (WSNs) gained increasing interests in recent years; since, they allow wide range of applications from environmental monitoring, to military and medical applications. As most of the sensor nodes (a.k.a. motes) are battery operated, they have limited lifetime, and user intervention is not feasible for most of the WSN applications. This study proposes a technique to reduce the energy dissipation of the processor component of the sensor nodes. We utilize a tiny cache-like structure called MoteCache between the CPU and the SRAM to cache the most recently used data values as well as to filter silent-store instruct . . .ions which write values that exactly match the values that are already stored at the memory address that is being written. A typical WSN application may sense and work on constant data values for long durations, when the environmental conditions are not changing rapidly. This common behavior of WSN applications considerably improves our energy savings. The optimal configuration of MoteCache reduces the total node energy by 24.7% on the average across a variety of simulated sensor benchmarks. The average lifetime of the nodes is also improved by 46% on the average for processor-intensive applications. Using the proposed technique, the lifetime of the nodes that run communication-intensive applications, such as TinyDB and Surge, is also improved as much as 14%. © Springer-Verlag Berlin Heidelberg 2006 Daha fazlası Daha az

Sertel, O. | Ünsalan, G.

Conference Object | 2006 | Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)4263 LNCS , pp.343 - 352

Segmenting 3D object surfaces is required for various high level computer vision and computer graphics applications. In computer vision, recognizing and estimating poses of 3D objects heavily depend on segmentation results. Similarly, physically meaningful segments of a 3D object may be useful in various computer graphics applications. Therefore, there are many segmentation algorithms proposed in the literature. Unfortunately, most of these algorithms can not perform reliably on freeform objects. In order to segment free-form objects, we introduce a novel method in this study. Different from previous segmentation methods, we first o . . .btain a function representation of the object surface in spherical coordinates. This representation allows detecting smooth edges on the object surface easily by a zero crossing edge detector. Edge detection results lead to segments of the object. We test our method on diverse free-form 3D objects and provide the segmentation results. © Springer-Verlag Berlin Heidelberg 2006 Daha fazlası Daha az

Kurt, B. | Gören, S.

Conference Object | 2018 | Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)10995 LNCS , pp.18 - 29

People spend a lot of time behind the wheel every day. Reading newspapers while driving a car is almost impossible. In this work, a mobile news reader application is developed to deliver the latest news from various sources to the drivers. The major difference from other news reader applications is that it is developed in accordance with the Ford SYNC technology. The user will be able to view the latest news on the SYNC screen while driving and listening to the selected news. In addition, drivers can select the desired news source and the desired news with the voice commands. Therefore, our proposed news reader application is an ena . . .bler for the drivers to follow the news in a safe way without distraction while keeping their hands on the wheel and their eyes on the road. © Springer International Publishing AG, part of Springer Nature 2018 Daha fazlası Daha az

Korkmaz, E.E.

Conference Object | 2006 | Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)4193 LNCS , pp.681 - 690

Linear Linkage Encoding (LLE) is a representational scheme proposed for Genetic Algorithms (GA). LLE is convenient to be used for grouping problems and it doesn't suffer from the redundancy problem that exists in classical encoding schemes. Any number of groups can be represented in a fixed length chromosome in this scheme. However, the length of the chromosome in LLE is determined by the number of elements to be grouped just like the other encoding schemes. This disadvantage becomes dominant when LLE is applied on large datasets and the encoding turns out to be an infeasible model. In this paper a twolevel approach is proposed for . . .LLE in order to overcome the problem. In this method, the large dataset is divided into a group of subsets. In the first phase of the process, the data in the subsets are grouped using LLE. Then these groups are used to obtain the final partitioning of the data in the second phase. The approach is tested on the clustering problem. Two considerably large datasets have been chosen for the experiments. It is not possible to obtain a satisfactory convergence with the straightforward application of LLE on these datasets. The method proposed can cluster the datasets with low error rates. © Springer-Verlag Berlin Heidelberg 2006 Daha fazlası Daha az