http://hdl.handle.net/10090/773 Search the Channel search http://dspace.nitle.org/simple-search http://hdl.handle.net/10090/6335 Title: Hex education: an introduction to computer hex <br/> <br/>Authors: Smith, Brendan <br/> <br/>Abstract: This thesis will explore the topic of computer Hex by taking an in-depth look at the game&#8217;s background, the ways in which artificial Hex players have been implemented to date, and some possible avenues for future advancement in this area. The first chapter provides an explanation of the rules of the game and some of the finer intricacies of Hex gameplay. The second chapter discusses a few of the mathematical properties of Hex that are of special importance to both gameplay and implementation. The third chapter recounts the history of the game, and the fourth chapter gives an overview of basic Hex strategy. An account of the first artificial Hex player, an analog machine, is given in chapter five. Chapter six discusses the standard techniques used to implement adversarial computer game players. Chapters seven and eight give a comprehensive analysis of two successful computer Hex players (Queenbee and Hexy), including their unique evaluation functions and search strategies. The final chapter compares these two Hex players, discusses their limitations, and takes a look at some possible new directions for computer Hex. <br/> <br/>Description: 57 leaves : ill. Includes bibliographical references. Thesis (B.A.)--Middlebury College, 2008. http://hdl.handle.net/10090/6229 Title: Primality testing <br/> <br/>Authors: Wehrwein, Jeff <br/> <br/>Abstract: Prime numbers have a very simple de&#12;finition, but they inspire an abundance of interesting and complex theory. Originally a topic of interest only to pure mathematicians, primes have found an important practical role in our modern technological world. This thesis introduces the &#12;eld of primality testing, with some mathematical background on the many interesting properties of prime numbers. The main focus of the thesis is the exploration of several diff&#11;erent primality tests from both an algorithmic and mathematical point of view. In addition, it explores some real-world applications of primality testing and discusses implementation issues and results. It concludes with a brief summary of current methods and possible directions of future improvements. <br/> <br/>Description: 87 leaves. Includes bibliographical references. Thesis (B.A.)--Middlebury College, 2008 http://hdl.handle.net/10090/802 Title: Parallelizing Numerical Integration <br/> <br/>Authors: Mitrevski, Petar <br/> <br/>Abstract: The ability to evaluate definite integrals is essential for virtually all fields of science. When a closed form expression cannot be obtained for a function, numerical integration may be used to approximate the value of the integral. There are various methods for numerical integration. In this study the rectangle, trapezoidal and Simpson&#8217;s rules were implemented into serial and parallel programs. All the programs were tested by integrating the functions by integrating the functions g(x) = &#8308;/&#8321;+ x&#8322;, h(x) = sin x/x, k(x) =3x&#8322; and f(x) = e - x&#8322;, between the limits: 10-9 and 1. The acquired integration results were compared to values obtained from Wolfram&#8217;s software package Mathematica. The running times of all the programs were recorded and compared for given integration parameters. As expected, the accuracy of the results obtained for a particular integration was found to be the greatest using Simpson&#8217;s rule and the lowest using the rectangle rule. When the number of segments was increased, the accuracy of the integration increased as well, however it was limited by the processor&#8217;s precision and the precision of the data type double in C. The running time of the programs on a single processor is constrained by the von Neumann bottleneck. By using parallel processors, it is possible to decrease the running time of the programs, while maintaining the accuracy of the results. In particular, doubling the number of processors was found to approximately halve the running time. The total error for a specific integration can be defined in terms of the integrated function and the number of segments used. The error scales with different powers of the number of segments used, depending on the integration rule. In case of the rectangle rule, a linear relationship was found between the error and the size of the integration segments. The error of the trapezoidal rule was proportional to the square of the segment size. Whereas, a quadric relationship was seen between the total error and the size of the segments when Simpson&#8217;s rule was applied. <br/> <br/>Description: Submitted in partial fulfillment of the requirement for the degree of Bachelor of Arts in Computer Science Middlebury College http://hdl.handle.net/10090/801 Title: A Comparison of Blocking & Non-Blocking Synchronization <br/> <br/>Authors: Chirls, Kevin <br/> <br/>Abstract: Synchronization is the problem of coordinating multiple events across time within a single system, such that the parallel events occur successfully without collisions. Computers use many different forms of synchronization for controlling hardware access, shared memory, shared files, process execution, as well as the handling of numerous other resources. It is especially important for data structures in a parallel processing environment. The status quo and most commonly used implementation of synchronization is blocking synchronization, which utilizes locks to prevent conflicts between competing processes. This thesis explores and compares the alternative of nonblocking synchronization, which does not employ locks, but instead complex algorithms using hardware-based primitive atomic operations. <br/> <br/>Description: Submitted in partial fulfillment of the requirement for the degree of Bachelor of Arts in Computer Science Middlebury College