CS Seminar Series

The Computer Science Department hosts Munster Technological University faculty and PhD students in a series of research seminars open to the department and university communities. If you are interested in presenting contact Dr Ruairí O'Reilly (ruairi.oreilly@mtu.ie) with a title, synopsis and bio.

Semester 2 Seminars - 2022

  • "Source Code Analysing and Inspection, Techniques and Technologies" by Dr Farshad Ghassemi Toosi - 04/03/22
  • "Human-Cyber-Physical Systems: A multi-disciplinary research area and strategy to target EU Horizon funding" by Dr Christian Beder - 04/03/22
  • "Neonatal seizure detection: A deep learning approach" by Dr Alison O’Shea - 18/02/22
  • "Addressing the Intra-class Mode Collapse Problem using Adaptive Input Image Normalization in GAN-based X-ray Images" by Mr Muhammad Muneeb Saad - 18/02/22
  • "Wireless communication for factory automation" by Dr Victor Cionca - 04/02/22
  • "Applications of Software-Defined Networking to Data Centre Network Management" Mr Jonathan Sherwin - 04/02/22

"Source Code Analysing and Inspection, Techniques and Technologies” – Dr Farshad Ghassemi Toosi

Abstract: Active and healthy Software systems always require maintenance. In some cases, up to 90% of the entire software budget is spent on maintenance. Maintenance includes running time optimization, bug detection and correction, dead-code detection (deleting discarded or obsoleted features), adding a new feature, changing the behaviour of an existing feature, measuring the software architecture divergence, feature location, and many other tasks.

The first step of software maintenance is to know the details of the software system structure, e.g., class/struct dependencies, method/function calls, variables/fields usages, nested blocks, naming conventions, pattern, and anti-pattern detection, comments and many other details specific to each language.

Source Code Analysing may be viewed as an umbral term that covers several different techniques and technologies specific to each language. The software system would be broken down into pieces and present a clearer and more understandable image of the software system where relations between software components are revealed and measured for code-smell, bug, anti-pattern detection and etc. Source-code analyzing may be performed statically or dynamically where static analysis does not require the system to execute and is more suitable for software systems with running limitations however the behavior of the system towards outside is not fully revealed using static analysis.

Bio: Farshad received his PhD from UL in 2017. He holds an MSc in Information Systems and BEng in Software Engineering. His main research interests include: Source Code Manipulation, Feature Location, Data analytics and the use of AI in Source Code analysing. He is a lecture in Computer Science Department at MTU (Cork) and a member of Lero (SFI research Centre).

"Human-Cyber-Physical Systems: A multi-disciplinary research area and strategy to target EU Horizon funding” – Dr Christian Beder

"Neonatal seizure detection: A deep learning approach" – Dr Alison O’Shea

Abstract: Detecting neonatal seizures is an important and time-sensitive task. The gold standard for neonatal seizure detection is through expert EEG analysis, but the advent of AI and machine learning have prompted the development of automated seizure detection algorithms. Neonatal EEG is a non-stationary, low amplitude signal and the presence of signal artifacts increases the challenge for seizure detection algorithms that are searching for evolving rhythmic patterns. This talk will discuss the development of feature-based machine learning algorithms for seizure detection and hopefully motivate our work in opting to research deep learning architectures.

Deep learning algorithms have shown state-of-the-art performance in both term and preterm seizure detection tasks (the architectures in this work have since been improved upon). Deep learning algorithms are perceived to be computationally intensive and require a large number of saved parameters, but a fully convolutional architecture trained to detect neonatal seizures has been implemented as part of an on-the-edge hand-held seizure detection device. Another characteristic of deep learning is the challenge in explaining and interpreting decisions, this work seeks to understand some aspects of the algorithm’s decision-making process and to utilise trained networks for EEG synthesis.

Bio: Alison received her PhD in Engineering from UCC in 2022 for her work with the Infant Research Centre. She holds an MSc. in technology entrepreneurship from the University of Notre Dame, Indiana, and a BEng. in electrical and electronic engineering from UCC. She has received recognition from Intel (Women in Technology Scholar 2010-2014), CEIA (Graduate of the Year 2014), and Google (Women Techmaker Scholar 2018) for her academic achievements. She worked with Qualcomm Ireland as part of their newly established Machine Learning team from 2018-2020. She joined the Computer Science Department in MTU in January 2021.

“Addressing the Intra-class Mode Collapse Problem using Adaptive Input Image Normalization in GAN-based X-ray Images” – Mr Muhammad Muneeb Saad

Abstract: Biomedical image datasets can be imbalanced due to the rarity of targeted diseases. Generative Adversarial Networks play a key role in addressing this imbalance by enabling the generation of synthetic images to augment and balance datasets. It is important to generate synthetic images that incorporate a diverse range of features such that they accurately represent the distribution of features present in the training imagery. Furthermore, the absence of diverse features in synthetic images can degrade the performance of machine learning classifiers. The mode collapse problem can impact a Generative Adversarial Network's capacity to generate diversified images. The mode collapse comes in two varieties; intra-class and inter-class. In this paper, the intra-class mode collapse problem is investigated, and its subsequent impact on the diversity of synthetic X-ray images is evaluated. This work contributes an empirical demonstration of the benefits of integrating the adaptive input-image normalization for the Deep Convolutional GAN to alleviate the intra-class mode collapse problem. Results demonstrate that the DCGAN with adaptive input-image normalization outperforms DCGAN with un-normalized X-ray images as evident by the superior diversity scores.

Bio: Muhammad Muneeb Saad received a bachelor’s degree from Gomal University Dera Ismail Khan, Pakistan (2016) and a Master’s degree from COMSATS University Islamabad, Abbottabad Campus, Pakistan (2018) in Electrical Engineering. Currently, he is working towards a Ph.D. degree at Munster Technological University (MTU), Ireland. Muneeb Saad is a recipient of the Risam scholarship award for his Ph.D. studies in Computing. His research interest includes artificial intelligence, deep learning, generative adversarial networks, and biomedical image analysis.

"Wireless communication for factory automation” – Dr Victor Cionca

Synopsis: Factory automation relies on control loops with very tight timing requirements. In such environments where robotic arms move at high speed and with high precision, missing deadlines or missing information can have disastrous consequences. Wireless communication is notoriously unreliable and it is difficult to provide guarantees. As part of the SFI CONFIRM project the team consisting of Dr. Alvaro de Medeiros as post-doctoral researcher, PhD students Mr. Yasantha Samarawickrama, and Ms. Tabinda Ashraf, led by myself as Funded Investigator, is developing solutions to make wireless communication suitable for factory automation. The primary hypothesis is that factory environments are somewhat predictable: they are governed by strict schedules calculated to optimise production. The ability to predict the characteristics of the environment can help scheduling and tuning the wireless communication to maximise the reliability and minimise the latency. Our work starts by designing methods to estimate and predict the status of the communication channel. Then we develop algorithms that, based on the channel state, optimally configure the parameters of the wireless transceivers. The final point that we address is link blockage: if a communication link is temporarily blocked there is very little that can be done in terms of adapting the physical parameters, the link is usually unusable for the length of the blockage. Instead we are exploring cooperative diversity where a transmission is processed by multiple receivers simultaneously, in effect going around the blockage. So far the work (parts of it) has been validated in simulation. In the near future we will be using Software Defined Radios to run experiments in the industrial workshops available on campus, as well as with industrial partners.

Bio: PhD from the University of Limerick in Wireless Sensor Networks and their security configuration. Postdoctoral positions in the WSN group in Tyndall National Institute, then in Nimbus centre. Research focus on wireless networks, short and long range, low and high power, from LoRa to mmwave, from the physical to the routing layers of the network stack. Interested in programmable networks, network control and optimisation, in-network intelligence.

“Applications of Software-Defined Networking to Data Centre Network Management” – Mr Jonathan Sherwin

Synopsis: Software-Defined Networking is the programmatic control of a network. It has been widely deployed in data centres, where it supports automated reconfiguration of networking devices in response to constantly changing traffic loads and user requirements. In this talk I will outline my work on SDN systems to assist a data centre operator to analyse the historical configuration and behaviour of their network.

Bio: Jonathan Sherwin (jonathan.sherwin@mtu.ie) received a B.Sc. degree in Computer Applications from Munster Technological University (MTU), Ireland in 1990, and an M.Sc. in Computer Science from University College Cork (UCC), Ireland in 2004. Following an internship in Apple, California in 1990, he worked as a Software Development Engineer and then Engineering Team Lead in Microsoft, Ireland until 1998. Since 1999, he has been a lecturer in the Department of Computer Science, MTU. He is currently pursuing a PhD in UCC in the area of Software-Defined Networking in Data Centre Networks. His research interests include network management, wireless networks, internet of things, and media streaming. He is a member of the ACM, and a Cisco-Certified Academic Instructor.

© 2024 CIT | web development by Granite Digital