From: 2023-12-18 09:15 to 15:00
Place: E:2405 (Glasburen)
Contact: birger [dot] swahn [at] cs [dot] lth [dot] se

Two Computer Science MSc theses to be presented on 18 December

Monday, 18 December there will be two master thesis presentations in Computer Science at Lund University, Faculty of Engineering.

The presentations will take place in E:2405 (Glasburen), one of them also online in zoom (please see link below).

Note to potential opponents: (Register as an opponent to the presentation of your choice by sending an email to the examiner for that presentation (firstname.lastname@cs.lth.se). Do not forget to specify the presentation you register for! Note that the number of opponents may be limited (often to two), so you might be forced to choose another presentation if you register too late. Registrations are individual, just as the oppositions are! More instructions are found on this page.)

09:15-10:00 in E:2405 (Glasburen)

Presenter: Hugo Bläckberg
Title: Optimizing Soak Test Reviews: A Comparative Study of Deep Learning Architectures
Examiner: Markus Borg
Supervisors: Patrik Edén (LTH), Gustav Lindroth (Neo4j), Eric Sporre (Neo4j)

Soak testing, a subset of system testing, aims to assess the long-term health of a system by running for an extended duration, such as several days. Soak testing aims to expose any performance degradations that occur over a longer time than other forms of system testing would. The manual review of results from a soak test makes it a time-consuming process. Investigating whether machine learning can be applied to identify anomalies in the results could reduce the time spent on review. We explored different deep learning architectures for this purpose, including long short-term memory autoencoders, transformers, and convolutional neural networks. The long short-term memory autoencoder and transformer classify anomalies based on reconstruction loss and a threshold value. In comparison, the convolutional neural network and transformer-encoder were trained with target labels. All models performed subpar, with low accuracies, except the cnn model with a balanced accuracy of 95% on the test set. Consequently, we propose a 1d convolutional neural network model that, with a high degree of accuracy, can classify a sub-sequence concerning a soak test as non-anomalous or anomalous.

Link to popular science summary: https://fileadmin.cs.lth.se/cs/Education/Examensarbete/Popsci/231218_09Bläckberg.pdf

14:15-15:00 in E:2405 (Glasburen) (also in zoom here: https://lu-se.zoom.us/j/66575215494)

Presenter: Andreas Karlsson
Title: Driving Development Resilience: Analyzing Truck Factors across Proprietary and Open-Source Projects
Examiner: Per Runeson
Supervisors: Markus Borg (LTH), Adam Tornhill (CodeScene AB) 

[Context] The agile approach to software development has led to developers retaining more project-specific knowledge. Additionally, the software industry faces a higher turnover rate compared to other sectors, making software projects more vulnerable to abrupt loss of key personnel.
[Objective] This paper investigates if the resilience of sudden loss differs in proprietary and open-source software projects. This is done through introduction of a new algorithm, expanding upon the current state-of-the-art Truck Factor algorithm, utilizing Lines of Code as heuristic.
[Method] The methodology comprises three phases. Phase one involves replicating the current state-of-the-art algorithm. In phase two, we introduce four new algorithmic approaches and analyze advantageous configurations. In phase three, we implement the most promising algorithm from phase two on open-source and proprietary projects obtained from CodeScene’s customers.
[Results] The difference between open-source and proprietary projects’ Truck Factor values are not significant. The majority of projects'’ has a low Truck Factor value, regardless of the context in which they are developed.
[Conclusion] Improving the distribution of knowledge within software organizations could enhance their resilience to sudden personnel loss. 

Link to popular science summary: https://fileadmin.cs.lth.se/cs/Education/Examensarbete/Popsci/231218_14Karlsson.pdf

Link to zoom presentation: https://lu-se.zoom.us/j/66575215494