Closed-Loop Data Science

Progress in sensing, computational power, storage and analytic tools has given us access to enormous amounts of complex data, which can inform us of better ways to manage our cities, run our companies or develop new medicines. However, the 'elephant in the room' is that when we act on that data we change the world, potentially invalidating the older data. Similarly, when monitoring living cities or companies, we are not able to run clean experiments on them - we get data which is affected by the way they are run today, which limits our ability to model these complex systems. We need ways to run ongoing experiments on such complex systems. We also need to support human interactions with large and complex data sets. In this project we will look at the overlap between the challenge someone faces when coping with all the choices associated with booking a flight for a weekend away, and an expert running complex experiments in a laboratory.
 
The project will test the core ideas in a number of areas, including personalisation of hearing aids, support for travel planning, analysis of cancer data, and adapting the computing resources for a major bank.
 

Partners

This project involves collaborative research with academic partners:  Glasgow Polyomics, the Urban Big Data Centre,  the University of Warwick, and industrial partners: Moodagent,   JP Morgan Technology & Operations, Glasgow, Skyscanner and Widex A/S, and is further supported by DataLab Scotland.
 

Vacancies

We are advertising the following 4 vacancies:

1 Post-Doctoral Research Associate/Fellow position in Computing Science on Data Systems position E20386  

To make a leading contribution to the EPSRC funded project “Closed-Loop Data Science for Complex, Computationally- and Data-Intensive Analytics” coordinated by PI Prof. Roderick Murray-Smith, working with Dr Nikos Ntarmos (line manager) and Dr Christos Anagnostopoulos.

Specifically, the job requires expert knowledge in large scale distributed systems, (big) data management and processing systems, and resource management and scheduling in distributed environments. The work will involve working closely with the technical team at JP Morgan’s Glasgow office, and there will be opportunities during the project for internships at JP Morgan.

The successful candidate will also be expected to contribute to the formulation and submission of research publications and research proposals as well as help manage and direct this complex and challenging project as opportunities allow.

Reference Number E20386
Location Gilmorehill Campus / Main Building
College / Service COLLEGE OF SCIENCE & ENGINEERING
Department SCHOOL OF COMPUTING SCIENCE
Job Family Research And Teaching
Position Type Full Time
Salary Range £35,210 - £39,610 / £43,266 - £50,132
https://www.jobs.ac.uk/job/BQZ341/research-associate-fellow
 
If you have interest in applying, please contact Roderick.Murray-Smith@glasgow.ac.uk
 
 
We have several Ph.D. opportunities in this area. If you are interested, contact Roderick.Murray-Smith@glasgow.ac.uk Current concrete Ph.D. opportunities include:
 
1.  Ph.D. scholarship (42 months, with stipend and fees for EU/UK citizens) on modelling and evaluating closed loop interactions in recommender systems supervised by Dr Craig Macdonald.  (Deadline 15th May 2019)

Machine learning techniques are widely used to address many recommendation scenarios – such as suggesting a movie to watch on (e.g.) Netflix, or recommending a point-of-interest to visit in a city, often by learning from historical user data. However, recommendation systems can be influenced by what users have already been recommended and thereafter viewed/visited, rather than what these systems might have found to be relevant of their own accord –for instance, Netflix might start to recommend movies that are already popular from its previous recommendations.

Such an effect can be described as a filter-bubble or a closed-loop feedback, and has been typically avoided through introducing novel or serendipitous recommendations into the suggestions. However, the alternative use of approaches originating from closed-loop theory, such as intermittent control, have not been systematically investigated within recommender systems.

This PhD will be focussed on applying ideas and techniques from closed-loop theory to state-of-the-art recommender systems. The candidate will investigate the modelling and deployment of closed-loop recommender systems using new neural networks architectures in comparison and along traditional matrix factorization and BPR-based recommenders. The evaluation of the resulting systems will be conducted using both public benchmarks in recommender systems as well as within the experimental pipeline of some of our data partners in the EPSRC Closed-Loop Data Science project.

The successful candidate will have a strong interest/background in recommender systems, machine learning, and/or information retrieval.

https://www.findaphd.com/phds/project/phd-in-computing-science-modelling-and-evaluating-closed-loop-interactions-in-recommender-systems/?p100765

 

2.  Ph.D. scholarship (42 months, with stipend and fees for EU/UK citizens) on Reinforcement learning in closed-loop data science. supervised by Professor Roderick Murray-Smith  and Dr Bjorn Jensen (Deadline 15th May 2019)

This studentship will explore reinforcement learning techniques in closed-loop data science. Progress in sensing, computational power, storage and analytic tools has given us access to enormous amounts of complex data, which can inform us of better ways to manage our cities, run our companies or develop new medicines. However, the 'elephant in the room' is that when we act on that data we change the world, potentially invalidating the older data. Similarly, when monitoring living cities or companies, we are not able to run clean experiments on them - we get data which is affected by the way they are run today, which limits our ability to model these complex systems. We need ways to run ongoing experiments on such complex systems. We also need to support human interactions with large and complex data sets. In this project we will look at the overlap between the challenge someone faces when coping with all the choices associated with booking a flight for a weekend away, and an expert running complex experiments in a laboratory.

Reinforcement learning is a key tool in adapting closed-loop system performance. The student will have the opportunity to apply techniques together with applications problems from  our research collaborators, which include Amazon, Moodagent, Skyscanner, JP Morgan and Widex.

 
3. PhD Scholarship (42 months with stipend and fees for EU/UK citizens) on  Moodagent - Conversational Music Assistant, Supervised by Jeff Dalton and Professor Roderick Murray-Smith  (Deadline 15th May 2019)

Conversational AI interfaces are beginning to allow people to communicate naturally with computers for the first time. This PhD, in partnership with Moodagent, focuses on intelligent music agents. At a basic level, this involves research and development of task-based music agents to allow people to interact with music (playing music, creating playlists, searching for music, etc…). Further, the goal is the development of an ‘intelligent’ agent that can incorporate contextual recommendations (music for running, music for a beach party, etc…) in a conversational context. For example, the context may include multiple users interacting (to create a playlist for a party) and the agent may help mediate the interactions. Going deeper, the aim is to develop conversational music intelligence that allows an agent to not simply perform tasks, but to converse with an agent about music. Music is a personal and subjective topic that is often vaguely defined, which make conversation in this domain both natural as well as an important research challenge. The explanatory descriptions from conversations about music will be used to explore research on deeper language-based understanding of music in order to improve quality and explainability of music retrieval and recommendation systems. This PhD will explore conversational music interactions in a cross-cultural and multilingual environment with Moodagent users in Denmark and India and will study key differences for this new and emerging group of users.

The PhD will study state-of-the-art task-based agent systems, including those based on deep learning models and that utilize reinforcement (and transfer) learning to learn agent policies. The research will also cover knowledge representation and research on construction of subjective personal knowledge graphs from conversation. It will incorporate conversational recommendation in multi-dimensional contexts, including social context, possibly involving groups of users.
 
 

Publications

Wandy, J., Davies, V., van der Hooft, J. J.J. , Weidt, S., Daly, R. and Rogers, S. (2019) In silico optimization of mass spectrometry fragmentation strategies in metabolomics. Metabolites, 9(10), 219. (doi:10.3390/metabo9100219) (PMID:31600991)

Ireland, D. G. , Doring, M., Glazier, D. I., Haidenbauer, J., Mai, M., Murray-Smith, R. and Ronchen, D. (2019) Kaon photoproduction and the Lambda decay parameter alpha. Physical Review Letters, (Accepted for Publication) Item availability restricted.

Jadidinejad, A. , Macdonald, C. and Ounis, I. (2019) How Sensitive is Recommendation Systems' Offline Evaluation to Popularity? In: REVEAL 2019 Workshop at RecSys, Copenhagen, Denmark, 20 Sep 2019, (Accepted for Publication) Item not currently available in Enlighten.

Davies, V. , Harvey, W. T., Reeve, R. and Husmeier, D. (2019) Improving the identification of antigenic sites in the H1N1 Influenza virus through accounting for the experimental structure in a sparse hierarchical Bayesian model. Journal of the Royal Statistical Society: Series C (Applied Statistics), 68(4), pp. 859-885. (doi:10.1111/rssc.12338)

Jadidinejad, A. , Macdonald, C. and Ounis, I. (2019) Unifying Explicit and Implicit Feedback for Rating Prediction and Ranking Recommendation Tasks. In: 5th ACM SIGIR International Conference on the Theory of Information Retrieval, Santa Clara, CA, USA, 02-05 Oct 2019, (Accepted for Publication) Item availability restricted.

Tonolini, F., Jensen, B. S. and Murray-Smith, R. (2019) Variational Sparse Coding. In: Conference on Uncertainty in Artificial Intelligence (UAI 2019), Tel Aviv, Israel, 22-25 July 2019,

Moran, O., Caramazza, P., Faccio, D. and Murray-Smith, R. (2018) Deep, Complex, Invertible Networks for Inversion of Transmission Effects in Multimode Optical Fibres. In: 32nd Conference on Neural Information Processing Systems (NeurIPS 2018), Montréal, Canada, 02-08 Dec 2018, Item not currently available in Enlighten.

Funded Projects

EPSRC funded project: £3M, 2018-2022:  Closed-Loop Data Science for Complex, Computationally- and Data-Intensive Analytics