Date: Wednesday 18th January 2012, 11:15, CSE102 Computer Science, Univ. of York


Speaker: Baoguo Yang, Department of Computer Science, Univ. of York

Topic: Community Detection in Networks (First Year Literature Review)

Detecting online communities is very significant in many areas,like biology, computer science and sociology where systems are usually viewed as graphs. In this talk, I will provide a brief introduction to this topic, and describe the existing representative methods as well as the testing algorithms, then I will list some future research directions and my suggested solutions.


Speaker: Suresh Manandhar, Department of Computer Science, Univ. of York

Topic: Dynamic and Static Prototype Vectors for Semantic Composition

In this talk, I will describe the work (jointly with Siva Reddy, Ioannis Klapaftis and Diana McCarthy) that I recently presented in IJCNLP-2011. The present work describes what are known as Compositional Distributional Semantic (CDS) models for computing the distributional semantics of a compound word given the distributional semantics of the component words. In conventional CDS models, a constituent word is typically represented by a single feature vector which has the disadvantage that it conflates word senses. However, it is also the case that not all the senses of a constituent word are relevant when composing the semantics of the compound. In this paper, we compare two different methods for selecting the relevant senses of constituent words. The first approach is based on Word Sense Induction and creates a static multi-prototype vectors representing the senses of a constituent word. The second approach creates a single dynamic prototype vector for each constituent word based on the distributional properties of the other constituents in the compound.

  

Date: Monday 23rd January 2012, 14:15, CSE102 Computer Science, Univ. of York

Speaker: Dr. Petr Kadlec, Evonik Industries AG, Germany (hosted by Daniel Kudenko)

Topic: Practical applications for predictive modeling with soft sensors

The presentation will deal with the application of predictive modelling in practical industrial settings. In the chemical industry this kind of models is known as "Soft Sensors". The soft sensors and their most commonly used types used in the chemical industry are introduced in the first part of the talk. As next, the challenges occurring when bringing the models from the test environment into real life operation are introduced. One the main challenges, namely the constantly changing environment and the consequences for the models, are discussed in more detail. The raised points are underlined by some practical applications within Evonik industries.

  

Date: Wednesday 25th January 2012, 11:15, CSE102 Computer Science, Univ. of York

Group Meeting

  

Date: Wednesday 1st Febuary 2012, 11:15, CSE102 Computer Science, Univ. of York

(no seminar - postponed to next week)

Speaker: Dimitar Kazakov, Department of Computer Science, Univ. of York

  

Date: Wednesday 8st Febuary 2012, 11:15, CSE102 Computer Science, Univ. of York

Speaker: Dimitar Kazakov, Department of Computer Science, Univ. of York

Topic: The Self-cognisant Robot

This is work in progress which discusses the challenge of developing self-cognisant articial intelligence systems, looking at the possible benefits and the main issues in this quest. It is argued that the degree of complexity, variation and specialisation of technological artefacts used nowadays, along with their sheer number, represent an issue that can and should be addressed through an important step towards greater autonomy, that is, the integration of learning, which will allow the artefact to observe its own functionality and build a model of itself. This model can be used to adjust the expectations from an imperfectly manufactured item, patch up its performance, and control its consistency over time, so providing a form of self- certication and a warning mechanism in case of deterioration. It is argued that these goals cannot be fully achieved without the ability of the learner to model its own performance, and the implications and issues of this self-reflective learning are debated. The article studies the link between learning about oneself, and the process of autopoiesis. A possible way of quantifying the faculty for self-cognition is proposed, and relevant areas of computer science, philosophy, and the study of the evolution of language are discussed.

  

Date: Wednesday 15th Febuary 2012, 11:15, CSE102 Computer Science, Univ. of York

Speaker: Mark Bartlett, Department of Computer Science, Univ. of York

Topic: Learning Family Trees from Genetic Data

Pedigrees (family trees) have many important applications in such wide-ranging areas as conservation, medical genetics and forensic science. If relationships are not known, the pedigree can be found through considering the genetic similarity of the individuals. The task becomes one of finding the maximum likelihood pedigree given the genetic data. However, the number of possible pedigrees is huge for even small numbers of individuals. This talk focusses on presenting an efficient method to find the most likely pedigree from this huge search space. Results show that the method outperforms the existing state-of-the-art approach.

  

Date: Wednesday 22th Febuary 2012, 11:15, CSE102 Computer Science, Univ. of York

Speaker: Alan Frisch, Department of Computer Science, Univ. of York

Topic: The Design of ESSENCE: A Constraint Language for Specifying Combinatorial Problems

This talk describes ESSENCE, a formal language for specifying combinatorial (decision or optimisation) problems at a high level of abstraction. ESSENCE is the result of our attempt to design a formal language that enables abstract problem specifications that are similar to rigorous specifications that use a mixture of natural language and discrete mathematics, such as those that appear in Garey and Johnson's catalog of NP-complete problems. ESSENCE, as well as this talk, is accessible to anyone with a background in discrete mathematics; no expertise in constraint programming is needed. The talk focuses on the abstraction features of ESSENCE, which enable problems to be specified directly and naturally.

  

Date: Monday 27th February 2012, 11:15, CSE102 Computer Science, Univ. of York


Speaker: Dhana Frerichs (final year undergraduate student - supervised by: Dimitar Kazakov)

Topic: Co-evolution of predator and prey behaviour

The study of co-evolution of competitive species has been the subject of much discussion amongst the biological community. This study will focus on the effect of co-evolution on the behaviour of species within a predator and prey relationship. Research into the subject of co-evolution will be used to help design and create a simulator in the form of a multi-agent system combined with a genetic algorithm. This simulator will then be used to conduct experiments using co-evolved and non-co-evolved populations of different generations to govern agent behaviour. These experiments will then provide information on how co-evolution affects the development of the agents involved compared to non-co-evolved agents.


Speaker: Priya Vasan (final year undergraduate student - supervised by: Dimitar Kazakov)

Topic: Equation Discovery for Financial Forecasting

Financial Forecasting is becoming a common practice in machine learning approaches and economic theories. Various methods have been developed to continually 'beat the market'. Several of these methods are discussed in the Literature Review chapter. This paper aims to select one method and further develops it to achieve better results than any other method. The equation discovery system, Lagramge, is the method that will be used in this paper. Lagramge takes input in the form of context free grammars and data sets, and returns a list of possible equations that are derived from the context free grammar, and whose parameters are based on the data set. The results of this computation are discussed and evaluated in later chapters of this paper.

  

Date: Wednesday 29th February 2012, 11:15, CSE102 Computer Science, Univ. of York

Speaker: Tommy Yuan, Department of Computer Science, Univ. of York

Topic: Safety argument review machine

Safety case development is not a post-development activity, rather it should occur throughout the system development lifecycle. The key components in a safety case are safety arguments. Too often, safety arguments are constructed without proper reasoning. Inappropriate reasoning in safety arguments could undermine a system's safety claims, which in turn contributes to safety-related failures of the system. To address this, we argue that informal logic argument schemes have important roles to play in safety arguments construction and review process. Ten commonly used reasoning schemes in computer system safety domain are proposed against the safety engineering literature. The role of informal logic dialogue games in computer system safety arguments reviewing is also discussed and a dialectical model for safety argument review is proposed. It is anticipated that this work will contribute toward the development of computer system safety arguments, and help to move forward the interplay between research in informal logic and research in computer system safety engineering.

  

Date: Wednesday 7th March 2012, 11:15, CSE102 Computer Science, Univ. of York

Group Meeting

  

Date: Wednesday 14th March 2012, 11:15, CSE102 Computer Science, Univ. of York

Speaker: William Smith, Department of Computer Science, Univ. of York

Topic: Face Appearance Modelling: Past, Present and Future

In this talk I will discuss attempts over the past 25 years to efficiently model the appearance of faces using statistical learning. This work all seeks to learn a low dimensional "facespace" from a sample of face images or 3D models. I will introduce classical techniques including Eigenfaces, Active Appearance Models and 3D Morphable Models and show how they have been used to help solve computer vision problems. I will then discuss how the face properties that these models seek to describe can be measured from imagery using computer vision techniques. I will present some of the work done in the CVPR group to advance the area of face modelling and model fitting. Finally, I will discuss my ideas for future work which seeks to describe faces at their most fundamental level: namely in terms of genetic and environmental subspaces.

  

Date: Wednesday 2nd May 2012, 11:15, CSE083 Computer Science, Univ. of York

Group Meeting

  

Date: Wednesday 9th May 2012, 11:15, CSE102 Computer Science, Univ. of York

(no seminar - postponed to 20th June)

Speaker: Jon Timmis, Department of Computer Science, Univ. of York

  

Date: Monday 14th May 2012, 14:00, CSE082 Computer Science, Univ. of York

Speaker: Ali Abusnina, Department of Computer Science, Univ. of York

Topic: Adaptive Soft Sensors (Literature Review)

Process monitoring and control systems procedures require building heavily instrumented plants on the account of cost. In worst cases, unavailability of the measuring instrument represents another obstacle other than the cost. These are few of the reasons that paved the path for Soft Sensors to proof themselves as a valuable alternative to traditional solutions. Soft Sensors makes it possible to keep the process fully monitored, and to measure hard-­‐to-­‐measure quantities. Such quantities are product quality, production efficiency, safety and other related measurements whose demand is continuously increasing in process industry in particular and in other engineering fields in general. Soft sensors have been considered for the last two decades to accomplish this task. However, the change in the process dynamics and in the environment the soft sensor is employed in necessitates the soft sensor developer to consider adaptivity. The presentation will give a modest overview on the machine learning techniques underlies Soft Sensors and details of Soft Sensors design

  

Date: Wednesday 16th May 2012, 11:15, CSE102 Computer Science, Univ. of York


Speaker: Sri Sri Perangur (final year undergraduate student - supervised by: Suresh Manandhar)

Topic: Virtual Actors: Machine emulation of human gesturing behavioral as portrayed by human actors

As the title describes it is focused on machine learning, human gesturing behavior. The project prototype was modeled after actress 'Julia Roberts' playing the role of 'Vivian' in the film 'Pretty Woman'. The project talk about : Quantifying environmental factors, Training data annotation, Machine learning system design for developing the classifiers, Machine learning cascading model for testing classifiers on unseen data (Note: both models implement supervised learning methods such as C4.5 algorithm for decision trees, Support Vector Machines 5 & 10 Fold), Application of the science to model characters of people we see in our every day life, From a Psychological perspective, and From a Sociology perspective.


Speaker: Alex Muller (final year undergraduate student - supervised by: James Cussens)

Topic: Allocating optional modules to University of York students using constrained optimisation

This final-year undergraduate project aimed to design and implement web-based software that could be used by departments and students to allocate optional modules more fairly and with less administrative overhead. Allocating modules "fairly" involves understanding how staff and students view fair allocation, and translating that into a constrained optimisation problem. This talk will discuss the way in which the optimisation problem was solved, the integration of the solver with the web application and the results from the application trial, run with the Archaeology and History departments.

  

Date: Wednesday 23rd May 2012, 11:15, CSE102 Computer Science, Univ. of York


Speaker: Kyriakos Efthymiadis, Department of Computer Science, Univ. of York

Topic: Overcoming Incorrect Knowledge in Plan-Based Reward Shaping (Practice talk for ALA Conference)

Reward shaping has been shown to significantly improve an agent's performance in reinforcement learning. Plan-based reward shaping is a successful approach in which a STRIPS plan is used in order to guide the agent to the optimal behaviour. However, if the provided knowledge is wrong, it has been shown the agent will take longer to learn the optimal policy. Previously, in some cases, it was better to ignore all prior knowledge despite it only being partially incorrect. This paper introduces a novel approach in overcoming incorrect domain knowledge when provided to an agent receiving plan-based reward shaping by the use of knowledge revision. Empirical results show that an agent using this method can outperform the previous agent receiving plan-based reward shaping without knowledge revision.


Speakers: Daniel Kudenko & Sam Devlin, Department of Computer Science, Univ. of York

Topic: Unidentified Research Object: Multi-Agent Knowledge-Based Reinforcement Learning

In this talk we will discuss approaches to modelling multi-agent specific knowledge and how they may be used to guide reinforcement learning. We will also discuss how the learning agents could then revise how we originally perceived the system and why having a human in the loop of knowledge revision/guidance may be beneficial.

  

Date: Wednesday 30th May 2012, 11:15, CSE102 Computer Science, Univ. of York


Speaker: Sam Devlin, Department of Computer Science, Univ. of York

Topic: Dynamic Potential-Based Reward Shaping (practice talk for AAMAS Conference)

Potential-based reward shaping can significantly improve the time needed to learn an optimal policy and, in multi-agent systems, the performance of the final joint-policy. It has been proven to not alter the optimal policy of an agent learning alone or the Nash equilibria of multiple agents learning together. However, a limitation of existing proofs is the assumption that the potential of a state does not change dynamically during the learning. This assumption often is broken, especially if the reward-shaping function is generated automatically. In this talk, I will prove and demonstrate a method of extending potential-based reward shaping to allow dynamic shaping and maintain the guarantees of policy invariance in the single-agent case and consistent Nash equilibria in the multi-agent case.


Speaker: Kleanthis Malialis, Department of Computer Science, Univ. of York

Topic: Reinforcement Learning of Throttling for DDoS Attack Response (practice talk for ALA Conference)

Distributed denial of service attacks (DDoS) constitute a serious and evolving threat in the current Internet. The most common type of these attacks is the flooding DDoS attack, which is designed to exhaust computer or network resources. Router throttling is a popular approach in the battle against these attacks, which views the flooding DDoS problem as a resource management or congestion problem. In this paper, we introduce a learning throttling approach which provides a highly adaptive response to such attacks. We compare our proposed approach against two other throttling approaches from the literature. It is shown that our approach effectively mitigates the impact of flooding DDoS attacks, and that it overcomes potential stability and convergence problems that the two throttling approaches suffer from.

  

Date: Wednesday 13th June 2012, 11:15, CSE102 Computer Science, Univ. of York

Speaker: Waleed Alsanie, Department of Computer Science, Univ. of York

Topic: Learning recursive PRISM programs with observed outcomes

PRISM is a probabilistic logic programming formalism which allows the combination of relational models and probabilistic models. This combination is desired to overcomes the limitations of both. This talk highlights a recent work on learning recursive PRISM programs with observed outcomes.

  

Date: Wednesday 20th June 2012, 11:15, CSE102 Computer Science, Univ. of York

Speaker: Jon Timmis, Department of Computer Science, Univ. of York

Topic: What can AI do for ....

In this talk I will discuss a number of research areas that I am working on, which have, what I hope you will think, are interesting challenges that might be of interest to people in the AI group. I will discus issues such as creating adaptive fault tolerant swarm robotic systems where we need intelligent approaches to the maintenance of swarm performance; creating swarm robotic systems that are safe for use with humans and coping with large amounts of data that needs to be processed on-line to detect different user behaviour (possible fraudulent use of a machine) and errors in system operation. I will provide high level overview of how we have gone about addressing each of these issues, and importantly what the problems are with our approach, and pose challenges hopefully in a way that will stimulate future collaboration.

  

Date: Wednesday 19th September 2012, 11:15, CSE102-103 Computer Science, Univ. of York

Speaker: Yann-Michaël De Hauwere, VU Brussels

Topic: Sparse Interactions in Multi-Agent Reinforcement Learning

Reinforcement learning has been used in unknown demains a high degree ef uncertainty. Also for domains in which multiple agents are acting together it is an interesting paradigm. This talk is concerned with settings where the interactions between agents are sparse. An efficient learning approach is to allow the agents to learn individually and only take the other agents into account when necessary. A key question is how to determine when interaction occurs. In this talk novel approaches are described which are capable of learning in which states such sparse interactions occur and based on this information use either a single agent approach or a multi-agent approach.

  

Date: Wednesday 24th October 2012, 11:15, CSE102 Computer Science, Univ. of York

Speaker: Daniel Whitehouse, Department of Computer Science, Univ. of York

Topic: Monte Carlo Tree Search with applications to Real-Time games and games with uncertainty

[David] will give an introduction to the Monte Carlo Tree Search (MCTS) family of algorithms. MCTS is a recent development in decision tree search which has led to huge advances in game AI (most notably computer Go) and many other applications. [His] presentation will include an overview of the MCTS algorithm and its many attractive features, as well as areas for future development. In addition [he] will talk about my PhD work on MCTS which is focused on the application of MCTS to games with hidden information and uncertainty. Finally, [he] will talk about our development of the winning entry to the 2012 PTSP competition held at WCCI and CIG (see http://www.ptsp-game.net/index.php), which uses a novel approach applying MCTS to a real-time game.

  

Date: Wednesday 31st October 2012, 11:15, CSE102 Computer Science, Univ. of York

Speaker: Alan Frisch, Department of Computer Science, Univ. of York

Topic: Combining Machine Learning and Optimisation (Unidentified Research Object)

[Alan] will first overview an ongoing research project at the 4C research lab that combines machine learning and optimisation. The goal of the work is to schedule work (say, on a production line or data centre) to take advantage of short-term variations in the price of energy. This is done by first using machine learning methods to predict how energy prices will vary and use optimisation methods to produce a schedule that saves on the cost of energy. Energy-aware scheduling can save millions of pounds a year. [Alan] will then open a discussion on how this work could be improved.

  

Date: Wednesday 7th November 2012, 11:15, CSE082 Computer Science, Univ. of York

Speaker: Philip Mourdjis, Department of Computer Science, Univ. of York

Topic: SpiNNaker and exascale computing

As the energy requirements for modern supercomputers meet practical limits a variety of new approaches are being developed in order reach exascale performance. The SpiNNaker chip multiprocessor (CMP) developed by the University of Manchester in association with ARM is a low power many core chip designed specifically for the modelling of spiking neural networks in the human brain. The presentation will begin with an overview of the aims for the spinnaker project before introducing the SpiNNaker CMP itself. The link to exascale computing will be explored with comparisons to alternative processing solutions made A discussion will attempt to identify additional applications for the SpiNNaker CMP.

  

Date: Wednesday 14th November 2012, 11:15, CSE102-103 Computer Science, Univ. of York

Group Meeting

  

Date: Wednesday 21st November 2012, 14:00, CSE082 Computer Science, Univ. of York

Speaker: Alan Frisch, Department of Computer Science, Univ. of York

Topic: A Decade of Research on Constraint Modelling and Reformulation: The Quest for Abstraction and Automation

(Departmental Seminar, a "Coffee-Table Book Seminar")
This talk reviews research in the field of constraint modelling and reformulation, focusing on the key themes of abstraction and automation. Looking to the future, the talk identifies key issues that must be confronted to further the quest for abstraction and automation. This is a high-level talk that assumes no background in constraint modelling.

  

Date: Wednesday 28th November 2012, 11:15, CSE102 Computer Science, Univ. of York

Speaker: Peter Cowling, Department of Computer Science, Univ. of York

Topic: Evolution of Fitness Functions to Improve Heuristic Optimization

The Variable Fitness Function approach evolves the weights of a weighted sum fitness function over the iterations of a search heuristic for a multiobjective optimization problem. The aim is to evolve a good direction for heuristic search, which makes use of the characteristics of the problem and of the search heuristic so as to perform better than the fixed search landscape given by the original weights. This is particularly true early in the search process when the original weights may provide little or no information about the future flexibility a orded by a given local search move. We apply the Variable Fitness Function to three very di erent case studies: the multiobjective Travelling Salesman Problem, the Virus 2-player zero-sum board game and a complex, real-world workforce scheduling problem. Variable Fitness Functions are evolved using a Genetic Algorithm which is shown to perform robustly across a wide range of parameter settings and problem instances. The Variable Fitness Function is found to improve heuristic performance for all our case studies. Generally evolving a Variable Fitness Function takes significant CPU time, but we show that after initial CPU-intensive training, evolved Variable Fitness Functions yield good performance on unseen problem instances, allowing significantly better solutions to be found for these problems without changing the search heuristic or using additional CPU time. Moreover, we show that for the Virus Game and the workforce scheduling problem, the variable fitness function has “learned” information that is transferable from instance to instance, although the precise nature of this information is di cult to understand.

  

Date: Wednesday 5th December 2012, 11:15, CSE102-103 Computer Science, Univ. of York

Speaker: Ole Torp Lassen, PLIS Research Group, Roskilde University

Topic: LoSt Highlights - PRISM for complex sequence analysis

Systems that combine logic programming and statistical inference allow machine learning systems to deal with both relational and statistical information. In the LoSt project, we experiment with applying one such system, PRISM (Taisuke Sato & Yoshitaka Kameya), to complex, large scale bio-informatical problems. As part of the general domain of DNA-annotation, the task of gene-finding is characterized by large sets of extremely long and highly ambiguous sequences of data, and represents a challenging setting for efficient analysis. We developed a compositional method, Bayesian Annotation Networks, where the complex overall task is optimized or approximated by identifying and negotiating interdependent constituent sub tasks and, in turn, integrating their analytical results according to their inter dependencies.

  

Date: Wednesday 12th December 2012, 11:15, CSE102-103 Computer Science, Univ. of York

Speaker: Haizhou (Joe) QU, Department of Computer Science, Univ. of York

Topic: Literature Review

  

Date: Thursday 13th December 2012, 15:30, CSE102-103 Computer Science, Univ. of York

Speaker: David Zendle, Department of Computer Science, Univ. of York

Topic: Literature Review

  

Last updated on 16 January 2013