About me:

I am a tenured Lecturer (Assistant Professor) in the School of Computer Science at NUI Galway.

Previously, I held appointments as:

  • Postdoctoral Researcher in the CIREGS lab at Cardiff University, UK. This work was funded by the EPSRC.
  • Postdoctoral Fellow at the Georgia Institute of Technology, USA within the ACES Research Group, in collaboration with Sandia National Laboratories. My work was funded by Sandia National Laboratories and the National Science Foundation.

Before this, I completed my PhD in Computer Science at the National University of Ireland, Galway. My PhD dissertation was nominated for the 2018 European AI PhD Dissertation award.

My research broadly falls under the heading of machine learning, but explores multiple topics including: neural networks, evolutionary computing, reinforcement learning, multi-agent systems and swarm intelligence. I’m also interested in applications of machine learning methods to problems including: renewable energy, smart homes, infrastructure planning, smart grid, robotics and cloud computing.

The research conducted within my research group is supported by funding from sources including: Science Foundation Ireland, Enterprise Ireland, and the Royal Irish Academy.

research_funding_logos2

Any potential MSc/PhD students who are interested in the topics outlined above and are looking for supervision should familiarize themselves with my previous work.

*Vacancies*: I currently have several openings for positions within my lab at undergraduate, postgraduate, and postdoctoral levels. Please get in contact with me if you are interested in completeing postdoctoral, PhD, or undergraduate research under my supervision. (Updated 29/03/2022)

Research profiles:   Google Scholar  ·  ResearchGate · Linkedin · NUIG

Panel 1

NEWS

News

13 May 2022: Congratulations to Junlin Lu, the first author of our most recent journal paper “A multi-objective multi-agent deep reinforcement learning approach to residential appliance scheduling” published in the Wiley journal IET Smart Grid today. Click here to read the paper.

11 May 2022: Excited to announce that I have been awarded an Science Foundation Ireland (SFI) Frontiers for the Future grant to support my upcoming project “Effective Integration of Renewable Energy within the Agriculture Sector in Ireland using Artificial Intelligence (EIRE AIAI)”. Thanks to NUIG and SFI for supporting this project. Also, congrats to my colleagues who were also awarded grants! SFI Press Release Link.

20 April 2022: I am happy to announce that I have openings for 1 Postdoctoral Researcher and 4 PhD students within my research group in the School of Computer Science at NUI Galway. You will work on an interdisciplinary project developing Artificial Intelligence (AI) algorithms and applying these AI algorithms to integrate renewable energy within the agriculture sector. See the following documents for more details. PhD vacancies: PhD_Add_NUIG_KarlMason. Postdoc vacancy: NUIG-RES-089-22-Job-Advert.

28 March 2022: We are accepting applications for the 2022 School of Computer Science Summer Research Scholarship at NUI Galway. Click here to find more information on this scholarship programme.

11 January 2022: The IMoFi – Intelligent Model Fidelity: Physics-Based Data-Driven Grid Modeling to Accelerate Accurate PV Integration Final Report has been released. Click here to view the final report.

01 July 2021: Our work was presented at the 2021 IEEE CEC conference today.

IEEE_CEC_2021_SlideImage

05 May 2021: Our Applied Energy journal paper ‘Investing in generation and storage capacity in a liberalised electricity market: an agent based approach’ is now available online: Click Here!

06 April 2021: Research with colleagues at the Georgia Institute of Technology has just been accepted for publication at the 2021 IEEE Congress on Evolutionary Computation (CEC) conference. Our paper is entitled ‘Building HVAC Control via Neural Networks and Natural Evolution Strategies’.

01 April 2021: Research with colleagues at Cardiff University has just been accepted for publication in the journal Applied Energy. Our paper is entitled ‘Investing in generation and storage capacity in a liberalised electricity market: an agent based approach’.

27 January 2021: Our Special Issue on Evolutionary Machine Learning in The Knowledge Engineering Review is now open for submissions. Myself and Dr. Patrick Mannion are guest editors for this Special Issue. More details can be found on the journal website: Link here. KER_journal

21 January 2021: Research with colleagues at the Georgia Institute of Technology has recently been accepted for publication at the 2021 IEEE Texas Power and Energy Conference (TPEC). Our paper is entitled ‘Detecting Behind-the-Meter PV Installation Using Convolutional Neural Networks’.

5th August 2019: The final published version of our review article ‘A review of reinforcement learning for autonomous building energy management’ is available at the following link: Click Here!

13th May 2019: I am delighted to have been nominated for the 2018 EurAI Artificial Intelligence Dissertation Award!

15th March 2019: A preprint of our new paper reviewing the applications of reinforcement learning for building energy management is available on arxiv! Link here.

Panel 2

Research Projects

Agent Based Modelling for Decision Making in the GB Energy Sector

This project focuses on using agent based modelling to study the strategic interactions between decision makers within the GB energy sector. There are multiple decision makers that influence the dynamics of energy systems, including: policy makers, energy generation companies and consumers. Understanding how these stakeholders interact with one another is key to understanding how the energy sector will evolve into the future.

ExampleDemandMapGB

Reinforcement Learning for Smart Buildings

The area of building energy management has received a significant amount of interest in recent years. This area is concerned with combining advancements in sensor technologies, communications and advanced control algorithms to optimize energy utilization. Reinforcement learning is one of the most prominent machine learning algorithms used for control problems and has had many successful applications in the area of building energy management. This project gives a comprehensive review of the literature relating to the application of reinforcement learning to developing autonomous building energy management systems and also focuses on developing new building energy management control algorithms.

smart-homeStockImage

 

Deep Learning and Solar Energy

I have conducted research relating to the application of Machine Learning methods such as convolutional neural networks to problems relating to solar energy. There is an ever increasing number of solar panels installed each year. There are a number of challenges associated with PV installation, including: merging solar into distribution feeders, estimating solar power production, etc. This project utilizes machine learning to address these challenges. This research was conducted in collaboration with Sandia National Laboratories.

solarPanelStockImage

 

 

Robot Maze Navigation and Novelty Search

Sensors

Novelty search is a recent approach to evolving neural networks that focuses on searching for networks with new and different behaviour rather than solely focusing on finding the network with the best objective fitness. In reality the concept of novelty is short lived in the sense that nothing stays new indefinitely. Algorithms that archive the best solutions to inform the search are therefore faced with the problem that the novelty scores of these archived solutions will change from generation to generation.

networkConfig

This project addresses this issue by proposing two methods of adjusting novelty scores of archived solutions: 1) Novelty Decay. 2) Recalculating Archived Novelty. Novelty decay enables novelty scores to decay overtime thus enabling the search algorithm to progress while recalculating novelty scores of the archived solutions updates the novelty of these solutions at each generation. When tested on the problem of maze navigation, it is observed that novelty decay and recalculating archived novelty converge faster than both objective search and novelty search alone.

MazesImg
Normal Maze (a) and Hard Maze (b). The robot must navigate from the blue marker to the target red marker.

 

Evolving Neural Networks

Evolutionary neural networks have proven to be a competitive approach for addressing a wide range of problems, from control to classification. The aims of this project were to: 1) Develop more effective and novel evolutionary neural network algorithms (often referred to as neuroevolution algorithms). 2) To utilize the effectiveness of these algorithms to address one of the most pressing problem that faces society, producing environmentally friendly energy. This contributions of this project were: 1) To conduct a meta optimisation analysis of the most prominent PSO velocity update equations. 2) The proposal of a novel method of evolving the topology and weights of a neural network, namely Neuro Differential Evolution (NDE) which implements a genetic algorithm to optimise the topology of the network and differential evolution to optimise the weights. 3) The proposal of a Multi-Objective Neural Network trained with Differential Evolution (MONNDE) algorithm to evolve networks that can produce Pareto fronts in dynamic multi-objective environments. 4) To apply these evolutionary neural networks to problems associated with energy systems and the environment, i.e. watershed management, multi objective dynamic economic emissions dispatch, forecasting wind power generation, power demand and CO2 emissions in Ireland, and finally forecasting CPU utilization in cloud computing data centers (accurate CPU forecasting is needed for efficient data center management, resulting in less energy consumption).

NDE_xover

NDE_mutation2

 

Swarm Intelligence Avoidance Strategies

This project focused on a well known swarm intelligence algorithm, Particle Swarm Optimisation (PSO) and explored how the performance of the algorithm can be enhanced by enabling particles to avoid the worst locations of the problem space. Typically in PSO, the algorithm functions as a result of the particles moving towards the best known locations (i.e. solutions to the optimisation problem).

 

This project proposed the PSO variant, PSO with birdsFlockingStockImageAvoidance of Worst Locations (PSO AWL). The benefit of avoiding the worst locations of the problem space is that particles waste less time exploring regions with poor solutions. This is particularly important when optimising problems with a high number of dimensions. The proposed PSO AWL performs well on both benchmark optimisation problems and engineering control problems (watershed management and power generator scheduling).

 

miscellaneous: Obstacle Avoiding Robot

Robotics is another area that I am interested in. I have built an obstacle avoiding robot (below) using an Arduino Uno micro controller. The robot detects objects using its sonar sensor which is fixed to a rotating mount. This allows the robot to look left and right when deciding what direction to move in.

RobotArduino1

 

miscellaneous: Java Poker Simulator

I am also interested in game playing and decision modelling. Here is a screenshot of a Texas Holdem Poker simulator that I created in java (still a work in progress).

Panel 3

Publications

Here is a short list of selected publications:

  • Mason, K., Qadrdan, M. and Jenkins, N., 2021. Investing in generation and storage capacity in a liberalised electricity market: An agent based approach. Applied Energy, 294, p.116905.
  • Mason, K., Reno, M.J., Blakely, L., Vejdan, S. and Grijalva, S., 2020. A deep neural network approach for behind-the-meter residential PV size, tilt and azimuth estimation. Solar Energy, 196, pp.260-269.
  • Mason, K. and Grijalva, S., 2019. A review of reinforcement learning for autonomous building energy management. Computers & Electrical Engineering, 78, pp.300-312.
  • Mason, K., Duggan, J. and Howley, E., 2018. Forecasting energy demand, wind generation and carbon dioxide emissions in Ireland using evolutionary neural networks. Energy, 155, pp.705-720.
  • Mason, K., Duggan, M., Barrett, E., Duggan, J. and Howley, E., 2018. Predicting host CPU utilization in the cloud using evolutionary neural networks. Future Generation Computer Systems, 86, pp.162-173.
  • Mason, K., Duggan, J. and Howley, E., 2018. A multi-objective neural network trained with differential evolution for dynamic economic emission dispatch. International Journal of Electrical Power & Energy Systems100, pp.201-221.

A full list of my publications can be found here.

Panel 4

Teaching and Supervision

 

Supervision:

Funded PhD Students:

  • Oct 2021 – 2025: Junlin Lu. Funder – CoSE.
  • Oct 2021 – 2025: Adam Callaghan. Funder – SFI CRT AI.
  • Oct 2021 – 2025: Sean Curley. Funder – CoSE.

Funded Research Assistants:

  • Sept 2021 – Nov 2021: Khadija Sitabkhan. Funder – Enterprise Ireland.

Other Funded Students:

  • 2021: Ethan Goodfellow. Funder – NUIG School of CS.

MSc Students:

  • 2021: Khadija Sitabkhan.
  • 2021: Eoin McAllister.
  • 2021: Oisin Doyle.
  • 2021: Junlin Lu.
  • 2021 – 2022: Oisin Brannock.
  • 2021 – 2022: Admir Olovcic.

Teaching:

CT5163 Applied Data Science with R.

CT103 Programming.

CT5181 Introduction to Artificial Intelligence.