lu.se

Datavetenskap

Lunds Tekniska Högskola

Kalendarium

CS MSc Thesis Presentations 7 June 2022

Föreläsning

Tid: 2022-06-07 09:15 till 16:00
Plats: E:2405 (Glasburen) and E:4130 (Lucas)
Kontakt: birger [dot] swahn [at] cs [dot] lth [dot] se
Spara händelsen till din kalender


Three Computer Science MSc theses to be presented on 7 June

Tuesday, 7 June there will be three master thesis presentations in Computer Science at Lund University, Faculty of Engineering.

The presentations will take place in E:4130 (Lucas) and E:2405 (Glasburen). See time and location for each presentation 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:4130 (Lucas)

Presenter: Emil Winzell
Title: X-ray Beamline Alignment Using Machine Learning at Max IV 
Examiner: Marcus Klang
Supervisors: Elin Anna Topp (LTH), Louisa Pickworth (Max IV)

The alignment process of beamlines at Max IV takes up a lot of the scientists time, which instead could be used for experiments. In the beamline, mirrors are modeled at micro scales in order to produce the correct light. This thesis explores possibilities and pitfalls in using machine learning for aligning the optical elements of a beamline. Three different approaches are evaluated: Variational autoencoders, reinforcement learning and bayesian optimization. The project contributes with valuable insights for future work of this complex problem.

Zoom link to presentation: https://lu-se.zoom.us/j/64981338167

Link to popular science summary: To be updated


11:00-12:00 in E:2405 (Glasburen)

Presenters: Maria Hellstrand, Anton Håkansson
Title: Increasing user engagement in an IoT system using a feedback monitoring system
Examiner: Alma Orucevic-Alagic​​​​​​​
Supervisors: Masoumeh Taromirad (LTH), Iraj Entezarjou (Inter IKEA), Laszlo Urszuly (Inter IKEA)

IoT has increased at a rapid pace and made its way into the everyday life of numerous homes. This puts a high demand on the IoT system and how information is provided to the users since a wide group of people needs to understand and utilize the system. This thesis investigates what kind of information is relevant and valuable for the users to receive. It can be frustrating to not understand why your system behaves in a certain way. This frustration could be avoided by providing users with the right feedback. It can also be beneficial for users to receive more information about the system and its health. By providing this information to the users, they can become more engaged with using their systems. Additionally, this thesis examines how information from the system can be collected and stored. Different approaches to acquiring this data are compared and discussed to choose the most feasible solution. The chosen solution in this thesis solves the stated problem by collecting valuable data from the internal communication channel in the system. The solution consists of two services that provide real-time feedback and an opportunity for the users to diagnose their systems. The results from this thesis show that by providing users with more information from their systems, their understanding and engagement with using the system can be increased.

Link to popular science summary: https://fileadmin.cs.lth.se/cs/Education/Examensarbete/Popsci/220607_11HellstrandHåkansson.pdf


15:15-16:00 in E:2405 (Glasburen)

Presenter: Patrik Persson
Title: Understanding Deep and Narrow Tree Search with GoExplore​​​​​​​
Examiner: Elin Anna Topp​​​​​​​
Supervisor: Volker Krueger (LTH)

The search space of large game trees can be reduced by restricting the depth or the breadth of the search. Most game-playing algorithms follow the first approach, where the tree is cut off at some depth and leaf nodes are evaluated according to an approximate value function that estimates the outcome of the game from this position. The approximated value function tends to be inaccurate in sparse reward environments, leading to poor overall performance. GoExplore avoids this problem by doing a deep and narrow search of the game tree, which has led to state-of-the-art results on sparse reward games such as Montezuma's Revenge and Pitfall. This thesis illustrates how GoExplore successfully drives this type of search by expanding the frontier towards novel states in the absence of rewards, and by re-focusing the frontier around rewards when they do occur.

Link to popular science summary: To be updated