Thursday 1 Dec, 1:15PM, room: G45
Deligiannis Pantazis: Algorithmic Skeletons - Towards a Heterogeneous Future
Synapsis:
Algorithmic skeletons, introduced by M. Cole in his PhD Thesis (1989), are high-level abstractionsthat enable the programmer to achieve parallelisation without having to worry about the associated complexities (e.g., synchronisation, coordination). In this talk I will present common skeletons,explain the advantages of skeletal programming (i.e., structured parallelism) and why they will play a critical role as we are heading towards heterogeneous devices (e.g., clusters of multicores, clouds, GPUs). Finally, I will present what I am working on since I started the PhD this September.
Dimitra Gkatzia: Natural Language Generation from Data
A lot of effort has been made to make machines talk like humans. Spoken dialogue systems, machine translation systems, summarization systems, recommendations systems and many more have been developed and used in several domains, e.g. medicine, weather forecasts, football commentaries, museum artifacts descriptions and others. All the aforementioned systems incorporate a Natural Language Generation (NLG) component in order to produce human language. However, NLG systems require a lot of developmental time and they are domain-specific, therefore their reusability is almost unfeasible.
In this talk, I will discuss about human language generation from data. The goal of these systems is to present the large amounts of unstructured and complex data in an understandable way for humans; human language. I will refer to the architecture of an NLG system, its components, different techniques used and some related systems in literature.
Tuesday 29 November 2011
Tuesday 15 November 2011
This weeks Psst! : Michael Kriegel: Crowd-sourced Story Creation, Vallejo, Marta:Urban Growth: Effects on Urban Green Spaces and implications for pla
Thursday 17 Nov, 1:15PM, room: G45
Michael Kriegel: Crowd-sourced Story Creation
My PhD work is motivated by a vision of interactive storytelling, where a computer system can "spin" and improvise a story in real-time taking into account a user's input and by the observation that it is incredibly hard to realize such a system. Part of the problem is the vast amounts of knowledge representation involved in describing a story world in enough detail that a program can reason about it. In my PhD I am investigating how crowd-sourcing might help to address this by asking 1) How can we extract coherent knowledge about the story world from many individual example stories and 2) what kind of infrastructure supports the crowd-sourced collection of this material.
Vallejo, Marta:Urban Growth: Effects on Urban Green Spaces and implications for planning and policy
Abstract:
The total amount of urban surface throughout the world represents only a 1-2%. However the resources needed to fulfill the necessities of the residents of these areas require the transformation of the 20% of the terrestrial surface above all in agricultural landscapes. The analysis of land use change spatial patterns is a crucial factor in order to understand ecological and social dynamics.
A key element in the study of urban evolution patterns is the peri-urban landscapes. A peri-urban landscape is a transition zone where coexists different land types as agricultural, forestry with urban residences, industry, transportation and leisure areas. In long term peri-urban areas are transformed rapidly in build-up zones and consequently new peri-urban areas are generated around these new built areas from former rural ones. This process should be controlled by nature conservation plans and green belt legislations due to the fact that speed in land-cover change is associated with rapid shifts in biodiversity.
The application of a range of different planning processes and regulatory policies to protect these landscapes can lead us to keep track of evolution of multiple hypothetical scenarios. The analysis of the plausible implications of each of them can give support to experts and politicians to understand the driving forces of the system and elaborate guidelines to mitigate possible negative effects in biota and human beings.
Michael Kriegel: Crowd-sourced Story Creation
My PhD work is motivated by a vision of interactive storytelling, where a computer system can "spin" and improvise a story in real-time taking into account a user's input and by the observation that it is incredibly hard to realize such a system. Part of the problem is the vast amounts of knowledge representation involved in describing a story world in enough detail that a program can reason about it. In my PhD I am investigating how crowd-sourcing might help to address this by asking 1) How can we extract coherent knowledge about the story world from many individual example stories and 2) what kind of infrastructure supports the crowd-sourced collection of this material.
Vallejo, Marta:Urban Growth: Effects on Urban Green Spaces and implications for planning and policy
Abstract:
The total amount of urban surface throughout the world represents only a 1-2%. However the resources needed to fulfill the necessities of the residents of these areas require the transformation of the 20% of the terrestrial surface above all in agricultural landscapes. The analysis of land use change spatial patterns is a crucial factor in order to understand ecological and social dynamics.
A key element in the study of urban evolution patterns is the peri-urban landscapes. A peri-urban landscape is a transition zone where coexists different land types as agricultural, forestry with urban residences, industry, transportation and leisure areas. In long term peri-urban areas are transformed rapidly in build-up zones and consequently new peri-urban areas are generated around these new built areas from former rural ones. This process should be controlled by nature conservation plans and green belt legislations due to the fact that speed in land-cover change is associated with rapid shifts in biodiversity.
The application of a range of different planning processes and regulatory policies to protect these landscapes can lead us to keep track of evolution of multiple hypothetical scenarios. The analysis of the plausible implications of each of them can give support to experts and politicians to understand the driving forces of the system and elaborate guidelines to mitigate possible negative effects in biota and human beings.
Wednesday 9 November 2011
Check out also: Maths and Statistic Seminars Blog
Maths seminars can be found here:
http://www.macs.hw.ac.uk/research/maths/seminars.htm
Maths PhD seminars are on Fridays 3:15 PM
http://www.macs.hw.ac.uk/research/maths/seminars.htm
Maths PhD seminars are on Fridays 3:15 PM
This weeks Psst! : Particle swarms categorizing documents, Dynamics of neural networks, and more
Thursday 10 Nov, 1:15PM, room: G45
Hamouda Khalifa H Chantar: Arabic Document Categorization using a combination of Binary Particle Swarm Optimization and K nearest neighbour.
Abstract—Document categorization is an important topic that is central to many applications that demand reasoning about and organisation of text documents, web pages, and so forth. Document classification is commonly achieved by choosing appropriate features (terms) and building a TFIDF document vector feature. In this process, feature selection is a key factor in determining the accuracy and effectiveness of resulting classifications. For a given classification task, the right choice of features means accurate classification with suitable levels of computational efficiency. Meanwhile, most document classification work is based on English language documents. In this work, we make three main contributions: (i) we demonstrate successful document classification in the context of Arabic documents (although previous work has demonstrated text classification in Arabic, the datasets used, and the experimental setup, have not been revealed); (ii) we offer our datasets to enable other researchers to compare directly with our results; (iii) we demonstrate a combination of Binary Particle Swarm Optimization and K nearest neighbour that performs well in selecting good sets of features for this task.
Gordon M Govan: Dynamics of neural networks with different motif distributions
In a network a 3-node motif is a group of three nodes linked together. The 3-node motif distribution indicates the percentages of 3-node motifs present in a network. We study the dynamics of Random Recurrent Neural Networks having two specific 3-node motif distributions. We find out that one of these motif distributions is more likely to have regular dynamics before any stimulus is applied and that few nodes need to be influenced in order to consistently have regular dynamics.
Majed Al-Saeed:TBA
Hamouda Khalifa H Chantar: Arabic Document Categorization using a combination of Binary Particle Swarm Optimization and K nearest neighbour.
Abstract—Document categorization is an important topic that is central to many applications that demand reasoning about and organisation of text documents, web pages, and so forth. Document classification is commonly achieved by choosing appropriate features (terms) and building a TFIDF document vector feature. In this process, feature selection is a key factor in determining the accuracy and effectiveness of resulting classifications. For a given classification task, the right choice of features means accurate classification with suitable levels of computational efficiency. Meanwhile, most document classification work is based on English language documents. In this work, we make three main contributions: (i) we demonstrate successful document classification in the context of Arabic documents (although previous work has demonstrated text classification in Arabic, the datasets used, and the experimental setup, have not been revealed); (ii) we offer our datasets to enable other researchers to compare directly with our results; (iii) we demonstrate a combination of Binary Particle Swarm Optimization and K nearest neighbour that performs well in selecting good sets of features for this task.
Gordon M Govan: Dynamics of neural networks with different motif distributions
In a network a 3-node motif is a group of three nodes linked together. The 3-node motif distribution indicates the percentages of 3-node motifs present in a network. We study the dynamics of Random Recurrent Neural Networks having two specific 3-node motif distributions. We find out that one of these motif distributions is more likely to have regular dynamics before any stimulus is applied and that few nodes need to be influenced in order to consistently have regular dynamics.
Majed Al-Saeed:TBA
Tuesday 1 November 2011
This weeks Psst! : Rich Pictures solve information mess + P2P hosts an eCommerce shop.
Thursday 3 Nov, 1:15PM, room: G45
Tessa Berg: Art as a Form of Enquiry
Information systems drive us all bonkers! How many times are we left hanging on the phone, stuck waiting in queues, sent wrong information, billed incorrectly or told ‘the computer says.. no’. They affect all of us on a daily basis. Why then, in this amazing technical world we live in, do we keep failing to get systems right?
I will endeavour to unwrap this mystery in 10 minutes!! The rich picture will be introduced as an effective tool for knowledge elicitation. Come along and I will explain my research area, what I am doing, how I am doing it and most importantly why I believe it to be significant to systems.
Ahmad Hadinata Fauzi: Community Trust Stores for Peer-to-Peer e-Commerce Applications
E-commerce applications have evolved from web-based selling via the Internet to selling in a P2P manner. P2P can enhance e-commerce applications to create lower cost systems compared to conventional client-server systems. However, P2P e-commerce applications will only be acceptable to users if they can provide robust, secure and equitable services to the peers involved during commercial transactions. In this paper, we propose use of a P2P shared store for trust information to support community based e-commerce applications. Nowadays, it can be economical and cheap to implement either in the cloud or in a distributed manner over the platforms of participating peers. Usage of a cheap and secure community store for trust data provides an effective alternative to conventional trusted third party support services for e-commerce transactions.
Fraser Blackmun: An Ongoing/Temporary Perspective in Social Network Analysis
Social Network Analysis (SNA) is a discipline that has explored people's networks of connections to one another for decades, using network-based measures and analytic techniques to identify and explain a social network's structure and properties. The modern rise of online social networking suggests a growing role for technology to mediate people's social connections. Looking ahead to the future, the concept of pervasive computing anticipates a world where technology is in a position to exploit real-world human interactions and bridge the physical/digital divide. New ways of thinking about social networks may be needed to help realise such a world, as well as to better understand the social world as it is now.
I am investigating a perspective in social network analysis that views any given social network as either 'Ongoing' or 'Temporary', which is the foundation of what I call the Ongoing/Temporary perspective. Mathematical models and assertion-based theories will be developed on Ongoing and Temporary networks that can work in harmony with traditional SNA methods to enrich analysis on any network. This perspective is intended to cast social networks in a new light, and to provide information about any given network's nature and likely future developments that would not necessarily be obvious using the SNA tools currently available.
Bamhdi, Alwi M: Mobile Ad Hoc Networks and Algorithms
Mobile Ad Hoc Networks have many applications, not least in 3rd world countries,disaster areas, and on the battlefield. There are many routing algorithms designed to effectively route packets through such networks. Our research is to investigate the effect of high density and high velocity on routing protocols by using NS2.
Tessa Berg: Art as a Form of Enquiry
Information systems drive us all bonkers! How many times are we left hanging on the phone, stuck waiting in queues, sent wrong information, billed incorrectly or told ‘the computer says.. no’. They affect all of us on a daily basis. Why then, in this amazing technical world we live in, do we keep failing to get systems right?
I will endeavour to unwrap this mystery in 10 minutes!! The rich picture will be introduced as an effective tool for knowledge elicitation. Come along and I will explain my research area, what I am doing, how I am doing it and most importantly why I believe it to be significant to systems.
Ahmad Hadinata Fauzi: Community Trust Stores for Peer-to-Peer e-Commerce Applications
E-commerce applications have evolved from web-based selling via the Internet to selling in a P2P manner. P2P can enhance e-commerce applications to create lower cost systems compared to conventional client-server systems. However, P2P e-commerce applications will only be acceptable to users if they can provide robust, secure and equitable services to the peers involved during commercial transactions. In this paper, we propose use of a P2P shared store for trust information to support community based e-commerce applications. Nowadays, it can be economical and cheap to implement either in the cloud or in a distributed manner over the platforms of participating peers. Usage of a cheap and secure community store for trust data provides an effective alternative to conventional trusted third party support services for e-commerce transactions.
Fraser Blackmun: An Ongoing/Temporary Perspective in Social Network Analysis
Social Network Analysis (SNA) is a discipline that has explored people's networks of connections to one another for decades, using network-based measures and analytic techniques to identify and explain a social network's structure and properties. The modern rise of online social networking suggests a growing role for technology to mediate people's social connections. Looking ahead to the future, the concept of pervasive computing anticipates a world where technology is in a position to exploit real-world human interactions and bridge the physical/digital divide. New ways of thinking about social networks may be needed to help realise such a world, as well as to better understand the social world as it is now.
I am investigating a perspective in social network analysis that views any given social network as either 'Ongoing' or 'Temporary', which is the foundation of what I call the Ongoing/Temporary perspective. Mathematical models and assertion-based theories will be developed on Ongoing and Temporary networks that can work in harmony with traditional SNA methods to enrich analysis on any network. This perspective is intended to cast social networks in a new light, and to provide information about any given network's nature and likely future developments that would not necessarily be obvious using the SNA tools currently available.
Bamhdi, Alwi M: Mobile Ad Hoc Networks and Algorithms
Mobile Ad Hoc Networks have many applications, not least in 3rd world countries,disaster areas, and on the battlefield. There are many routing algorithms designed to effectively route packets through such networks. Our research is to investigate the effect of high density and high velocity on routing protocols by using NS2.
Subscribe to:
Posts (Atom)