Major gains have been made in recent years in object recognition due to advances in deep convolutional neural networks. One struggle with deep learning is identifying an optimal network architecture for a given problem. Often different configurations are tried until one is identified that gives acceptable results. This paper proposes an asynchronous learning algorithm that finds an optimal network configuration by automatically adjusting network hyperparameters.

In this paper, we study a challenging problem of how to pool multiple ride-share trip requests in real time under an uncertain environment. The goals are better performance metrics of efficiency and acceptable satisfaction of riders. To solve the problem effectively, an objective function that compromises the benefits and losses of dynamic ridesharing service is proposed. The Polar Coordinates based Ride-Matching strategy (PCRM) that can adapt to the satisfaction of riders on board is also addressed. In the experiment, large scale data sets from New York City (NYC) are applied. We do a case study to identify the best set of parameters of the dynamic ridesharing service with a training set of 135,252 trip requests. In addition, we also use a testing set containing 427,799 trip requests and two state-of-the-art approaches as baselines to estimate the effectiveness of our method. The experimental results show that on average 38% of traveling distance can be saved, nearly 100% of passengers can be served and each rider only spends an additional 3.8 minutes in ridesharing trips compared to single rider service.

Weighted voting games are important in multiagent systems because of their usage in automated decision making. However, they are not immune from the vulnerability of false‐name manipulation by strategic agents that may be present in the games. False‐name manipulation involves an agent splitting its weight among several false identities in anticipation of power increase. Previous works have considered false‐name manipulation using the well‐known Shapley–Shubik and Banzhaf power indices. Bounds on the extent of power that a manipulator may gain exist when it splits into k = 2 false identities for both the Shapley–Shubik and Banzhaf indices. The bounds when an agent splits into k > 2 false identities, until now, have remained open for the two indices. This article answers this open problem by providing four nontrivial bounds when an agent splits into k > 2 false identities for the two indices. Furthermore, we propose a new bound on the extent of power that a manipulator may gain when it splits into several false identities in a class of games referred to as excess unanimity weighted voting games. Finally, we complement our theoretical results with empirical evaluation. Results from our experiments confirm the existence of beneficial splits into several false identities for the two indices, and also establish that splitting into more than two false identities is qualitatively different than the previously known splitting into exactly two false identities.

Interaction between agents is one of the key factors in multiagent societies. Using interaction, agents communicate with each other and cooperatively execute complex tasks that are beyond the capability of a single agent. Cooperatively executing tasks may endanger the success of an agent if it attempts to cooperate with peers that are not proficient or reliable. Therefore, agents need to have an evaluation mechanism to select peers for cooperation. Trust is one of the measures commonly used to evaluate the effectiveness of agents in cooperative societies. Since all interactions are subject to uncertainty, the risk behavior of agents as a contextual factor needs to be taken into account in decision making. In this research, we propose the concept of adaptive risk and agent strategy along with an algorithm that helps agents make decisions in an self-adaptive society utilizing an agent’s own experience and recommendation-based trust. Trust-based decision making increases the profit of the system along with lower task failure in comparison to a no-trust model in which agents do not utilize evaluation mechanisms for choosing their cooperation peers.

Human pose tracking is among the most popular hotspots in the field of computer vision. In this paper, we propose a novel game theory based method for tracking two dimensional articulated human poses in monocular video sequences. A new probability scheme of game theory is introduced into human pose tracking to find optimal solutions of human poses. The possible limb positions are modeled as strategies of agents who play normal form game with adjacent agents. Likelihood measurements and distance constraints are applied to calculate the payoffs of each of the strategies. Finally, the Nash equilibria are found for each normal form game and the human poses are estimated based on them. In the experiments, the effectiveness and efficiency of the proposed algorithm is fully exhibited.

The number of undergraduates entering computer science has declined in recent years. This is paralleled by a drop in the number of high school students taking the CS AP exam and the number of high schools offering computer science courses. The declines come at a time when career opportunities in CS continue to grow and computer science graduates are seen as crucial in building a globally competitive workforce for the 21st century. Efforts aimed at reversing the declining interest in computer science include curriculum revisions at the undergraduate level at many institutions, a re-design of computer science AP courses [1], and the inclusion of computational thinking into disciplines outside computer science [3].

While high-performance architectures have included some Instruction-Level Parallelism (ILP) for at least 25 years, recent computer designs have exploited ILP to a significant degree. Although a local scheduler is not sufficient for generation of excellent ILP code, it is necessary as many global scheduling and software pipelining techniques rely on a local scheduler. Global scheduling techniques are well-documented, yet practical discussions of local schedulers are notable in their absence. This paper strives to remedy that disparity by describing a list scheduling framework and several important practical details that, taken together, allow implementation of an efficient local instruction scheduler that is easily retargetable for ILP architectures. The foundation of our machine-independent instruction scheduler is a timing model that allows easy retargetability to a wide range of architectures. In addition to describing how a general list-scheduler can be implemented within the framework of our timing model, experimental results indicate that lookahead scheduling can profoundly improve a scheduler's ability to produce a legal schedule. Further experimental data shows that deciding to schedule a data dependence DAG (DDD) in forward or reverse order depends significantly upon that target architecture, suggesting the possibility of scheduling in each direction and using the best of the two schedules. In contrast, experiments demonstrate little difference in code quality for schedules generated by either instruction-driven or operation-driven schedulers. Thus, the inherent flexibility of operation-driven methods suggests including that approach in a retargetable instruction scheduler. List scheduling is, of course, a heuristic scheduling method. A variety of scheduling heuristics are presented. In addition, the paper describes a method, using a genetic algorithm search, to ‘fine-tune’ the weights of twenty-four individual heuristics to form a DDD-node heuristic tuned to a specific architecture. © 1998 John Wiley & Sons, Ltd.

Software pipelining is a technique that reforms the loop to improve execution time. Iterations are executed in overlapped fashion to increase parallelism. Modulo scheduling places each operation so that the schedule is legal when replicated and offset by a target initiation interval. This process is repeated with larger initiation intervals until success is achieved. Kernel recognition methods schedule operations as rapidly as possible until a pattern is recognized. These two distinctly different methods have various strengths and weaknesses. This paper explores the benefits and draw-backs of each.

Article Free Access Share on Teaching computer science concepts and problem solving with a spreadsheet Authors: Mary Veronica Kolesar Department of Computer Science, Utah State University, Logan, Utah Department of Computer Science, Utah State University, Logan, UtahView Profile , Vicki H. Allan Department of Computer Science, Utah State University, Logan, Utah Department of Computer Science, Utah State University, Logan, UtahView Profile Authors Info & Claims SIGCSE '95: Proceedings of the twenty-sixth SIGCSE technical symposium on Computer science educationMarch 1995 Pages 10–13https://doi.org/10.1145/199688.199698Published:15 March 1995Publication History 10citation434DownloadsMetricsTotal Citations10Total Downloads434Last 12 Months29Last 6 weeks4 Get Citation AlertsNew Citation Alert added!This alert has been successfully added and will be sent to:You will be notified whenever a record that you have chosen has been cited.To manage your alert preferences, click on the button below.Manage my AlertsNew Citation Alert!Please log in to your account Save to BinderSave to BinderCreate a New BinderNameCancelCreateExport CitationPublisher SiteeReaderPDF

Anti-dependencies are a major cause of bottleneck in software pipelining. Anti-dependencies can be removed in software by code duplication as in the technique of modulo variable expansion. However, it is possible to eliminate anti-dependencies in hardware. So far, two such VLIW architectures have been proposed: polycyclic and URPR-1 architectures. These architectures also introduce attractive solutions for the memory contention problem in general. We, on the other hand, propose a more cost-effective architecture for the same purpose: SRFA. SRFA is based on "shifting register files".>

When data dependency graph arcs representing data dependency information are annotated with minimum and maximum timing information, new algorithms are required. Foresighted compaction is a list scheduling technique in which look ahead is used in making decisions. Foresighted compaction is very effective in reducing, failure inherent in greedy compaction algorithms.>

Software Pipelining is a fine-grain loop optimization technique for architectures that support synchronous parallel execution.We compare Lam's software pipelining algorithm with Ebcio~lu and Nakatani's technique.This research seems to indicate the Enhanced Pipeline Scheduling algorithm is a good general purpose software pipelining algorithm, due to the fact that it performs only slightly worse than Lam's algorithm on single basic block loops and should perform better than Lam's algorithm on multiple basic block loops.However, if pipelining single basic block loops is the goal, it appears that it would be better to use Lam's algorithm.We s3so propose a technique for changing the resource constrained scheduling priority of operations to prevent operations from future iterations from being significantly delayed due to resource confllcts.

In current microcode generation systems, one simplification that is frequently made is to assume an absence of timing restrictions. It is critical that timing is considered when the target architecture involves branch delays, volatile registers, or microoperations requiring multiple microinstructions to complete. A general form for representing synchronous timing in clocked microarchitectures and methods of compacting data-dependency graphs with general timing are described.>

We consider the complexity of the problem of encoding microoperations into the fields of a control word on a horizontally microprogrammed machine, and show it to be NP-complete. The major result of this correspondence is a formalization of the microinstruction encoding problem, which makes the complexity argument straightforward.