We would like to invite all the PhD students in ECS to give a short presentation about 
their PhD research at the machine learning and the game theory reading groups (especially if your research is somewhat related to machine learning or game theory).

This is an opportunity for you to receive feedback about your research and practice with delivering short talks - which often occurs in conferences. 

Format

10 mins slot for each student: 5mins presentation + 5mins discussions.

Schedule

- Machine learning session: 2 April 2015, 1:00pm
- Game theory session: 26 March 2015, 1:00pm
Note: we will schedule more sessions if needed.

Requirements

There are no strict requirements for this talk. If you have just started and you've never given a talk about your PhD, it is fine to present preliminary work or ideas you are currently investigating. Instead, if you feel that the group is already familiar about your work, it is fine to focus on more specific aspects of your research or on a paper that you have recently published or submitted.

Submit your interest

If you wish to accept this invitation, please just send an email (with a rough title/abstract, if you have) to:

-Machine learning: Dr Matteo Venanzi (mv1g10@ecs.soton.ac.uk)
-Game theory: Dr Dengji Zhao (d.zhao@soton.ac.uk)

We look forward to your reply!

Machine Learning Session 

Date: 2 April, 1:15pm - 2:00pm

Alexandros Zenonos (13:15pm – 13:25pm)

Title: Coordinating Measurements for Environmental Monitoring in Participatory Sensing Settings
Abstract: Environmental monitoring is important, as it allows authorities to understand the impact of potentially harmful environmental phenomena such as air pollution, noise or temperature, on public health. To achieve this effectively, participatory sensing is a promising paradigm for large-scale data collection. In this approach, ordinary citizens (non-expert contributors) collect environmental data using low-cost mobile devices. However, these participants are generally self-interested agents having their own goals and making local decisions about where and when to take measurements, if any at all. This can lead to a highly inefficient outcome, where observations are either taken redundantly or do not provide sufficient information about key areas of interest. To address these challenges a coordination system is necessary to guide and to coordinate participants.

Elliot Salisbury (13:30pm – 13:40pm)

Title: Crowd-controlled UAVs
Abstract: Finding and tracking targets and events in a live video feed is important for many commercial applications, from CCTV surveillance used by police and security firms, to the rapid mapping of events from aerial imagery. However, understanding and interpreting this live video in a way that benefits the end user, requires an understanding of natural language descriptions of a given target, and the comprehension of context within a video stream that cannot currently be automated and instead requires human supervision. The current state of artificial intelligence, especially vision, is still inadequate to meet these requirements. Hence, in this paper, we consider the use of real-time crowdsourcing to identify and track targets given by a natural language description. In particular we present a novel method for harnessing real-time crowd input for live video tagging, and a novel method of tailoring the tasks to individual workers to increase response rate from the crowd workers.

James Holyhead (13:45pm – 13:55pm)

Title: Consumer Targeting in Residential Demand Response Programmes
Abstract: Demand response refers to a family of techniques that are available to electricity suppliers to aid with balancing supply and demand, typically by calling on consumers of electricity to reduce consumption during periods of high demand. In this paper we propose a novel approach to residential demand response, in which incentives are targeted at the subset of consumers who are both relevant (likely to use shiftable appliances, such as washing machines and dishwashers during peak hours) and willing to reduce (likely to react positively to a reduction request from their electricity supplier). To this end, we present a mixed integer programming solution that finds the optimal subset of consumers to target with incentives. We show that our solution is capable of significantly reducing supplier costs and smoothing peaks in electricity demand by targeting only a subset of the consumer pool.

Game Theory Session

Date: 26 March, 1:15pm - 2:00pm

Laurie M. Carver (1:15pm - 1:25pm)

Title: Stability and Efficiency of Interbank Lending Equilibria under Systemic Risk Taxes
Abstract: To prevent state bail-outs of banking systems as occurred during the crisis of 2007-9, taxes on banks have been proposed to pay for any in the future. Two such are the ‘Tobin’ financial transactions tax and more recently a Pigou-style tax based on financial network theory. By analysing an interbank lending model, the former can be shown to be inefficient, while a version of the latter is efficient. However, practical considerations mean that its proposed implementation could destabilise the system’s equilibrium and reduce efficiency below a zero tax environment. This is termed the ‘price of planning’, in analogy to the price of anarchy.

Beining Shang (1:25pm - 1:35pm)

Title: Behavioural Sorting for Swarm Roboots with Hardware Variations
Abstract: Swarm robotic systems can offer advantages of robustness, flexibility and scalability, just like social insects. One of the issues that
researchers are facing is the hardware variation when implementing real
robotic swarms. Identical software can not guarantee identical behaviours among all robots due to hardware differences between swarm members. We propose a novel approach for sorting swarm robots according to their hardware differences. This method is based on the large number of interactions between robots and the environment. Individual robot’s unique hardware circumstance determines its unique decision and reaction during each robotic controlling step, and these unique local reactions accumulate and contribute to the robot’s global behaviour. Accordingly by separating these hardware-triggered global behaviours, swarm robots can be sorted according to their hardware variations.

Radu Pruna (1:35pm - 1:45pm)

Title: A New Structural Stochastic Volatility Model of Asset Pricing
Abstract: The Fundamentalists vs Chartists model design is highly motivated by observations made on how real financial traders behave. Empirical evidence shows that, by large, these are the two clusters of forecasting behaviour in financial markets. Most of the currently existing Agent-Based Computational Finance Models focus on generating data that matches a series of statistical properties widely observed in real life. We extend a Structural Stochastic Volatility Model of Asset Pricing by implementing a more complex Fundamentalist vs Chartist model. By increasing the number of strategies used by Chartists and making them more complex, we would like to demonstrate that the new model preserves a series of stylized facts obtained by previous ones while also revealing new properties. From the system’s perspective we are able to offer a clear view of how external information changes the market behaviour and to what extent.

Xue (Sherry) Yang (1:45pm - 1:55pm)

Title: Video Advertising in Real Time bidding
Abstract: Real Time bidding (RTB) is a new way to buy and sell digital advertising. Unlike traditional digital sales—where ad inventory is sold in blocks, with pre-arranged commitments—RTB creates a unique transaction for every single impression at the moment the ad is shown. I will introduce how a video advertisement present to a user’s browser and what is the problem exist in this online advertising system.

Comments