• Refine Query
  • Source
  • Publication year
  • to
  • Language
  • 6857
  • 2288
  • 1
  • 1
  • Tagged with
  • 9147
  • 9121
  • 8130
  • 8071
  • 1264
  • 925
  • 898
  • 703
  • 668
  • 661
  • 626
  • 552
  • 460
  • 426
  • 360
  • About
  • The Global ETD Search service is a free service for researchers to find electronic theses and dissertations. This service is provided by the Networked Digital Library of Theses and Dissertations.
    Our metadata is collected from universities around the world. If you manage a university/consortium/country archive and want to be added, details can be found on the NDLTD website.
931

Analysis and comparison of solving algorithms for sudoku : Focusing on efficiency

Ekne, Samuel, Gylleus, Karl January 2015 (has links)
The number puzzle sudoku has been steadily increasing in popularity. As the puzzle becomes more popular, so does the demand to solve it with algorithms. To meet this demand a number of different sudoku algorithms have been developed. This report will examine the most popular algorithms and compare them in terms of efficiency when dealing with a large number of test cases.
932

Factorization patterns in 3-SAT : Analysis of clause ordering for reductions of the integer factorization problem

Häggvik, Adrian, Khalili, Marwan January 2015 (has links)
Dividing a large integer into factors is a well-known difficult problem that is heavily used in modern cryptography. One approach to achieve faster factorizations is to reduce the Integer factorization problem to another difficult problem, for example the Boolean satisfiability problem. Many studies have been done on the optimization of algorithms that can solve the Boolean satisfiability problem effectively and multiple programs have been created for the purpose of achieving efficiency. This study focuses on one of these programs, MiniSAT, with the objective of finding patterns in the input that result in faster runtimes for solving reduced instances of the Integer factorization problem. The method used for finding these patterns was to create different reductions from the Integer factorization problem to the Boolean satisfiability problem and then solving the reduced instances using MiniSAT. The order of the clauses in the boolean formulas were changed in order to analyze which order yields the best results. In total, seven orderings were tested for factorizing four products using three different reductions. Our conclusions are that none of the patterns lead to better results for every test case. Not changing the order of the clauses will achieve consistent results, however there are cases when shuffles with variance could be preferable. / Uppdelning av ett stort heltal i faktorer är ett välkänt svårt problem som vardagligen används i modern kryptografi för att kryptera data. Ett visst sätt att uppnå snabbare tider av faktoriseringen går ut på att reducera instanser av faktoriseringsproblemet till ett annat svårt problem, till exempel satisfierbarhetsproblemet. Många studier har gjorts med avseende på optimeringar av algoritmer som löser satisfierbarhetsproblemet, och flera program har skapats i syftet att lösa problemet mer effektivt. Denna studie fokuserar på ett av dessa program, MiniSAT, med målet att hitta mönster i indatan som resulterar i snabbare lösningstider för reducerade instanser av faktoriseringsproblemet. Metoden som används för att hitta dessa mönster var att skapa olika reduktioner från faktoriseringsproblemet till satisfierbarhetsproblemet och sedan lösa de reducerade probleminstanserna med MiniSAT. Ordningen av klausulerna i de booleanska formlerna ändrades och resultaten för varje ordning analyserades. Totalt testades sju olika sorteringar för att faktorisera fyra produkter med tre olika reduktioner. Våra slutsatser är att inget specifikt mönster ger det bästa resultat för varje testfall. Att inte ändra ordningen på klausulerna ger konsistenta resultat, men det finns fall då en ordning som har varierande resultat kan vara att föredra.
933

Smirking or Smiling Smileys? : Evaluating the Use of Emoticons to Determine Sentimental Mood

Lousseief, Elias, Hindersson, Tobias January 2015 (has links)
Machine Learning classifiers are commonly used for the purpose of Sentiment Analysis. These classifiers use annotated training data from which they learn to predict the sentiment of texts, for example whether a text conveys a positive or a negative sentiment. In this thesis we compare the performance of two sources of training data for the purposes of sentiment classification on Twitter: (i) tweets annotated by hand of a fixed quantity (about 2000 tweets) and (ii) tweets annotated automatically by an emoticon heuristic of increasing quantity (from 2000 tweets to 1.6 million tweets). The performance of these training sets are evaluated by training commonly used classifiers (Naive Bayes, Support Vector Machines and Maximum Entropy) and comparing the classification accuracy of the different data sets on a test set annotated by hand. These tests are made with varying use of n-gram models (unigrams, bigrams, and a combination of both) and the varying use of a stop word filter. We show that while the hand-annotated training set performs well in equally sized training sets, the automatically annotated training set exceeds the accuracy of the hand-annotated training set in all test setups but one when 1.6 million automatically annotated tweets are used for training. / Maskininlärningsalgoritmer används ofta för att utföra analys av känslomassig inställning; sentimentsanalys. Dessa algoritmer använder annoterad träningsdata för att lära sig att klassificera texter efter exempelvis huruvida de speglar ett positivt eller negativt sentiment. I den här uppsatsen företas sentimentsanalys av data från Twitter varvid effektiviteten utvärderas med avseende på två typer av träningsdata: (i) en fix mängd tweets som annoterats för hand (cirka 2000 tweets) och (ii) olika mängder tweets som genomgått automatisk annotering av en heuristik baserad på emoticons (från 2000 till 1.6 miljoner tweets). Effektiviteten som träningsdata hos dessa dataset har utvärderats genom att träna vanliga maskininlärningsalgoritmer (Naive Bayes, Support Vector Machines och Maximum Entropy) vartefter jämförelser gjorts av hur väl de lyckats klassificera ett set med testdata som annoterats för hand. Testerna har gjorts med olika typer av n-gram (unigram, bigram samt kombinationen av dessa) samt valbar inkludering av ett filter med stoppord. I studien framkommer att träningsdata annoterad för hand presterar bra i jämförelse med annoteringar som gjorts heuristiskt förutsatt att dataseten är av samma storlek. Då omfattningen av den heuristiskt annoterade träningsdatan växer förbättras dock förmågan till korrekta klassificeringar, och när storleken uppgår till 1.6 miljoner tweets ger användning av handannoterad träningsdata bättre resultat i endast ett fall av de testupptällningar som använts.
934

Factoring integers with parallel SAT solvers

Lundén, Daniel, Forsblom, Erik January 2015 (has links)
Factoring integers is a well known problem that at present cannot be solved in polynomial time. Therefore, other approaches for solving factorization problems are of interest. One such approach is to reduce factorization to SAT and solve it with a dedicated SAT solver. In this study, parallel SAT solvers are examined in this context and evaluated in terms of speedup, efficiency and effectiveness versus sequential SAT solvers. The methodology used was an experimental approach where different parallel and sequential solvers were benchmarked on different reductions from integer factorization to SAT. The benchmarks concluded that parallel SAT solvers are not better suited for solving factorization problems than sequential solvers. The performance boosts achieved by the fastest parallel solver barely corresponded to the extra amount of available parallel resources over the fastest sequential solver. / Att faktorisera heltal är ett välkänt problem som för närvarande inte kan lösas i polynomisk tid. Därför är andra tillvägagångssätt för att lösa faktorisering av intresse. Ett sådant tillvägagångssätt är att reducera faktorisering till SAT och sedan lösa problemet med en dedikerad SAT-lösare. I denna studie undersöks parallella SAT-lösare i detta sammanhang och utvärderas i förhållande till uppsnabbning, effektivitet och ändamålsenlighet jämfört med sekventiella SAT-lösare. Den metod som användes var en experimentell sådan där olika parallella och sekventiella lösare jämfördes på olika reduktioner från heltalsfaktorisering till SAT. Genom testerna erhölls slutsatsen att parallella SAT-lösare inte är bättre lämpade för att lösa heltalsfaktorisering än sekventiella lösare. Prestandavinsterna som uppnåddes av den snabbaste parallella lösaren motsvarade knappt den extra mängd parallella resurser som denna hade över den snabbaste sekventiella lösaren.
935

Simulating Infants’ Word-to-Object Mapping : The impact of focusing attention to relevant objects and reducing noise when learning language

Magnusson, Sebastian January 2015 (has links)
This project used a computer program to simulate how infants may learn new languages. The project drew inspiration from a report written by Jeffrey Mark Siskind in 1997, in which he described some main problems that infants face when learning a new language. The experiments conducted in this project relate to recent discoveries by Aimee E. Stahl and Lisa Feigenson suggesting that infants show an ability to focus their learning to objects in their surroundings that may be especially relevant at the moment. The computer program was then run with different settings to see how its ability to learn new words was affected by how many objects it was presented with when hearing a word and how often the objects presented were irrelevant to the word heard. Both factors affected the learning rate significantly. / I detta projekt användes ett datorprogram för att simulera hur spädbarn kan lära sig nya språk. Projektet hämtade inspiration från en rapport skriven av Jeffrey Mark Siskind 1997 där han beskrev några huvudsakliga problem som spädbarn möter när de lär sig ett nytt språk. Experimenten som utfördes i detta projekt relaterar till nya rön från Aimee E. Stahl och Lisa Feigenson som tyder på att spädbarn har en förmåga att fokusera sitt lärande till objekt i deras omgivning som kan vara särskilt relevanta för tillfället. Datorprogrammet kördes sedan med olika inställningar för att se hur dess förmåga att lära sig nya ord påverkades av hur många objekt det fick presenterat för sig när det hörde ett ord och hur ofta de presenterade objekten inte hade med det hörda ordet att göra. Båda faktorerna hade signifikant effekt på hur snabbt inlärningen skedde.
936

Scheduling of Modern Elevators : A description of modern elevators and a comparison of heuristics used for scheduling them

Björkholm, Viktor, Bränn, Jesper January 2015 (has links)
Three heuristics for scheduling of a modern elevator system are evaluated and compared to one another. Modern elevators are described as an elevator system with multiple cabins per elevator shaft with the capability of travelling horizontally as well as vertically in a unidirectional system. / Tre heuristiker för schemaläggning av moderna hissystem är utvärderade och jämförda med varandra. Moderna hissystem beskrivs som hissystem med fler än en hissvagn per schakt och som har kapaciteten att färdas både horizontellt och vertikalt.
937

Game Theory in Social Media with Quantal Response Equilibrium

Elmgren, Rasmus, Blomquist, Eric January 2015 (has links)
This paper examines the possibility to construct a Game Theory model to describe Social Media with a Quantal Response Equilibrium. It is based on a literature study. The paper is influenced by "A Game-theoretic Model of Attention in Social Networks" written by Goel and Ronaghi but creates a more realistic model by replacing their Nash Equilibrium with a Quantal Response Equilibrium. Such model is constructed in the Result section and elaborated further in the Discussion. This paper also discusses the difficulties of Game Theory in Social Media and the flaws of the model created in the Result. The model helps provide an understanding of success in Social Media. It is possible to do continued research with more emphasis on the value of different players or how the order of content affects the level of attention.
938

Automated scheduling : Performance in different scenarios

Sjöberg, Jonas, Nissar, Hugo January 2015 (has links)
Scheduling is and has always been a time consuming problem. While our everyday life becomes more automated, more things need to be scheduled. This report presents established algorithms for automated scheduling by conducting a basic study of relevant literature, while testing the speed of the algorithms for different scenarios. The study shows that from the most commonly used algorithms, the forward chaining is the fastest in most scenarios but will not always find a solution. A partial-order planner will always find the solution if there is one, but with the expense of time. The conclusion is that the partial-order algorithm performs better overall when reliability is more important than speed.
939

A study of algorithms for 2-dimensional self-assembly in swarm robotics

Wong, Bonny, Boldt-Christmas, Axel January 2015 (has links)
Intelligent Control is a novel way to solve problems involving the control of complex systems in many areas. Among these areas are multi-robot systems. Swarm robotics is a multi-robot system that is inspired by natural swarms such as ones found among social insects. One of the problems found within swarm robotics involves the forming of geometrical shapes in 2D-space through local interactions. This thesis has taken an existing solution to the shape forming problem and addressed some of the limitations of it. Through the use of a computer simulation, a swarm robotic system using these improvements were simulated and yielded positive results.
940

Distributed Graph Mining : A study of performance advantages in distributed data mining paradigms when processing graphs using PageRank on a single node cluster

Abdlwafa, Alan, Edman, Henrik January 2015 (has links)
Distributed data mining is a relatively new area within computer science that is steadily growing, emerging from the demands of being able to gather and process various distributed data by utilising clusters. This report presents the properties of graph structured data and what paradigms to use for efficiently processing the data type, based on comprehensive theoretical studies applied on practical tests performed on a single node cluster. The results in the study showcase the various performance aspects of processing graph data, using different open source paradigm frameworks and amount of shards used on input. A conclusion to be drawn from this study is that there are no real performance advantages to using distributed data mining paradigms specifically developed for graph data on single machines.

Page generated in 0.0569 seconds