Feldman PhD Scholarship in Computational Statistics
Applications are invited for a PhD funding opportunity to conduct research in an area at the interface between Statistics and Computer Science, commencing in September 2016.
The studentship is tenable for 3 years and covers tuition fees at the UK/EU rate plus a stipend of £16,539 per annum (based on the standard UK Research Council rate with London weighting). International students may also apply, but will need to source additional funding to meet the difference in cost between UK/EU and overseas tuition fees. The procedure for determining one’s tuition fee status is outlined on the UCL website.
The successful candidate will be based either in the Department of Computer Science or the Department of Statistical Scienceand will have one supervisor from each.
Studentship Description
The following two projects are available (outline descriptions are linked to the project titles).
- Learning representations with latent variable models
Supervisors
Dr David Barber (Computer Science)
Dr Ricardo Silva (Statistical Science)
Outline
There has been an explosion of research in highly non-linear large-scale data analysis. Currently most large-scale latent variable deep learning models are rather generic with latent variables used to simply increase the power of the distributional representation. In contrast, the project would study highly structured latent variable models and associated inference algorithms with the aim to marry probabilistic modelling with large scale non-linear processing. In particular, our proposed focus is on the study of sequential and network data: given data on decisions performed by intelligent agents, interactions among components in a social or engineered system, or environmental states that change over time, which unobserved components explain the patterns seen? The goals are to predict future system behaviour and to cluster different patterns of activities. Methods include the combination of non-linear representations given by neural networks with algorithms for approximate inference in stochastic systems. Possible applications include models of human navigation and change point detection in networks, among others.
- Multi-agent deep reinforcement learning for machine bidding in digital advertising
Supervisors
Dr Jun Wang (Computer Science)
Dr Jinghao Xue (Statistical Science)
Outline
Display advertising has become a significant battlefield for research on big data and machine learning. It is evident that the transaction volume (over 100 billion auction traded daily) from display advertising has already surpassed that of the financial market. As the advertising transactions are aggregated across websites in real time, the display advertising industry has a unique opportunity to understand the internet traffic, user behaviour, and online transactions.
Teamed up with experts in both Computer Science and Statistical Science, in this PhD project we will aim to formulate novel multi-agent reinforcement learning for machine bidding in ad exchange. Our study will be focused on reinforcement learning in a second price auction context, studying the learners (as advertisers) involved in the bidding of an impression. We use Microeconomic models such as auction theory and game theory to model different party’s incentives and their objectives to make a prediction about the market price from the historical data and competition, and investigate how all this information can be fed in and integrated with a deep reinforcement learning framework. The solution will have wide application, ranging from online advertising, e-commerce, finance and other fields which require the combination of online prediction and modelling user incentives.
Alternatively, candidates may propose their own topics, provided that they have already been discussed with prospective supervisors in both departments.
Person Specification
For admission to the MPhil/PhD Statistical Science, the requirement is a first or high upper second class Bachelor’s degree, or a Master’s degree with merit or distinction, in Mathematics, Statistics, Computer Science, or a related quantitative discipline. Overseas qualifications of an equivalent standard are also acceptable. Further details can be found on the Departmental website.
For admission to the MPhil/PhD Computer Science, the normal requirement is a Bachelor’s degree with first or upper second-class Honours, and/or a distinction at Master’s level, in an appropriate subject. Applicants with other qualifications and sufficient relevant experience and background knowledge may also be considered. More information, including English Language qualifications can be found on the Departmental website.
How to Apply
All new candidates should apply initially for the Research Degree: Statistical Science (RRDSTASING01) by completing the online form and, in addition (and very importantly), send a separate covering letter making their case for the funding. The covering letter should be sent to Dr Russell Evans at stats.pgr-admissions@ucl.ac.uk. Candidates already in receipt of an offer of admission (to either department) need only send the covering letter.
This opportunity has expired. It was originally published here:
https://www.ucl.ac.uk/statistics/prospective-postgraduates/dept-studentships/feldman2