lu.se

Datavetenskap

Lunds Tekniska Högskola

Denna sida på svenska This page in English

Five Computer Science Master Thesis Seminars

2014-03-14

March 14th was the day for the third coordinated master thesis presentations in Computer Science at Lund University, Faculty of engineering.

—————————————

kl 09.15: Filling In Safety Impact Analysis Forms (Quality, Efficiency and Soft Issues)

Författare/Authors: Johan Thiborg-Ericson

Examinator/Examiner: Lars Bendix (LTH)

Abstract: 
A company wanted to improve the correctness, quality, efficiency and handling of a impact analysis form filled in by developers as as a part of their maintenance process. Sometimes answers were missing, inconsistent or ironic because of the layout and disagreement about when to add regression tests. Critizising the form was concidered dangerous by the developers. Four handling problems were found: the developers copied old forms, they implemented bug fixes before the CCB had allowed them to do so and worked with many forms simultaneously. Also, the review of the forms was too long after the form had been filled in. The company should have a policy on how product quality precautions and efficiency should be balanced or they should assign representatives for the developers. The question with ironic answers should be removed. The time between writing and reviewing the form should be shorter.

—————————————


kl 10.15: Going beyond textual information in defect duplicate detection: An exploratory study in the Android issue tracker


Författare/Authors: Jens Johansson

Handledare/Supervisor: Markus Borg (LTH)

Examinator/Examiner: Per Runeson (LTH)

Abstract: 
In a large software project different users and testers may submit defect reports corresponding to the same defect, i.e. duplicates are created in the bug tracking repository. In this thesis defects extracted from the Android bug tracking system is used in order to measure how duplicates can be detected using Lucene's information retrieval engine which measures textual similarities. Also a logistic regression is performed to examine the possibility to predict if a defect report is a duplicate of another based on previously known textual similarity measures between duplicates. The results shows that a majority of the known duplicates in the corpus was possible to retrieve in a top ten search result list using Lucene's information retrieval engine. While the logistic regression approach did not yield many false positives, a significant amount of the known duplicates were not detected using this technique.

—————————————


kl 13.15: Comparison of power consumption in OpenGL ES 2.0/3.0

Författare/Authors: Filip Karlsson

Handledare/Supervisor: Andreas Alptun (Illusion Labs)

Examinator/Examiner: Michael Doggett (LTH)

Abstract: 
Power efficiency is an important property to software, especially on mobile devices. This report compares the power consumption in two different ways of rendering particles with OpenGL ES 2.0/3.0. The first approach uses vertex buffer objects and the second method uses transform feedback buffers that were introduced in OpenGL ES 3.0 (although not new to desktop versions of OpenGL). All tests are performed on an iPhone 5S. The report also discusses what unit is appropriate to use when measuring the power consumption of particle systems..

—————————————


kl 14.15: Model- and Case-Based Decision Support in Planning


Författare/Authors: Juri Pyykkö

Handledare/Supervisor: Jacek Malec (LTH)

Examinator/Examiner: Klas Nilsson (LTH)

Abstract: 
There are many potential benefits of decision support in mission planning, and in this work a problem solving technique known as case-based reasoning was applied in order to achieve this. It uses knowledge from previous problems to solve new problems in accordance with the heuristic that similar problems have similar solutions. To find similar problems old and new problem descriptions are compared, but this comparison is complicated by the fact that they differ in composition. These difficulties served as the focus of this work, where several methods of coping with the same were examined and evaluated. Additionally, the problem descriptions were encapsulated by regular objects in the object-oriented programming sense of the word, which are static entities once compiled, but in order to make use of the decision support, they would necessarily have to be modified at run-time, to which end a technology known as ontology was utilized, and furthermore examined as well.

—————————————


kl 15.15: Algorithmic Engineering Aspects of Fast Zeta Transform-based Graph Colouring Algorithms.

Författare/Authors: Mats Rydberg

Handledare/Supervisor: Thore Husfeldt (LTH)

Examinator/Examiner: Jonas Skeppstedt (LTH)

Abstract: 

The chromatic polynomial G(t) of a graph G on n vertices is a univariate polynomial of degree n, passing through the points (q; P(G; q)) where P(G; q) is the number of q-colourings of G. In this paper, we present an implementation of an algorithm by Björklund, Husfeldt, Kaski and Koivisto that computes G(t) in time O(2n) and space O(1:2916n). We compare the performance of two different core libraries to eachother and show our performance against an implementation done by Haggard, Pearce and Royle from 2010. We also present the chromatic polynomials for a small Queen graph and a certain graph specified by Hillar and Windfeldt.



Exjob@cs