CS Other Presentations

Department of Computer Science - University of Cyprus

Besides Colloquiums, the Department of Computer Science at the University of Cyprus also holds Other Presentations (Research Seminars, PhD Defenses, Short Term Courses, Demonstrations, etc.). These presentations are given by scientists who aim to present preliminary results of their research work and/or other technical material. Other Presentations serve as a forum for educating Computer Science students and related announcements are disseminated to the Department of Computer Science (i.e., the csall list):

RSS Directions

Presentations Coordinator: Demetris Zeinalipour

Seminar (PENEK Research Results): Understanding Self-Control through Computational Modelling of Internal Conflict, Aristodemos Cleanthous (University of Cyprus, Cyprus), Tuesday, November 30th, 2010, 11:30-12:30 EET.

The Department of Computer Science at the University of Cyprus cordially invites you to the Seminar (PENEK Research Results) entitled:

Understanding Self-Control through Computational Modelling of Internal Conflict

Abstract:
This thesis proposes a novel computational model of internal conflict which aims to provide further understanding on this highly complex and perplexing condition of the human brain. In particular, the purpose of this thesis is to identify specific factors which influence and enable internal conflict to be resolved by self-control behaviour. Individuals are likely to experience an internal conflict when evaluating the same outcomes of choice along distinct dimensions or criteria. A value conflict of this sort can be resolved as if it was a result of strategic interaction between rational subagents of the brain. The particular setting for this interaction is a well-studied theoretical game, the Iterated Prisoners Dilemma, where the mutual cooperation outcome of the game corresponds to the behaviour of self-control. The computational system developed for the purposes of this thesis realises this particular view of internal conflict by implementing two spiking neural networks as two agents competing in the Iterated Prisoners Dilemma, where the agents pursue individual value maximisation through simultaneous but independent learning. This high-level game theoretical approach to the problem of internal conflict incorporates at the same time biological realism through the employed neuronal model, the process of learning, as well as by relating the agents and their actions in the game with particular brain regions and their functioning. In particular, the spiking neural networks comprise of leaky integrate-and-fire neurons, while the learning process is implemented by reinforcement of stochastic synaptic transmission as well as by reward-modulated spike-timing-dependent plasticity with eligibility trace. Moreover, the action of cooperation and defection by each agent maps to a greater relative activity of fronto-parietal and limbic system areas respectively. As demonstrated through numerous simulations, the artificial neuronal system behaved efficiently in the game theoretical framework because the learning agents implemented the optimum result for the system through consistent mutual cooperation. Therefore self-control behaviour can indeed be learned (since it corresponds to mutual cooperation), and as showed by further results, it is enhanced by strong reward-correlated memory. Moreover, the ability of the agents to adopt optimal counter strategies as a response to their competitors, enabled the identification of particular value structures that characterise internal conflicts of low and high intensity that promote or hinder the attainment of self-control behaviour. In the process of obtaining the results which are relevant to the problem of self-control behaviour and internal conflict, this thesis work applied for the first time spiking neural agents combined with biological plausible reinforcement learning in a highly demanding multiagent task. In addition, further results with our system showed that high firing irregularity at high rates enhances learning.

Short Bio:
Aristodemos Cleanthous received his B.Sc in Mathematics and Economics at the London School of Economics and Political Science (LSE) in 2002. He further studied at the University College London (UCL), receiving his M.Sc in Computer Science in 2005. In 2006 he started his PhD degree at Department of Computer Science, University of Cyprus and on 29 October 2010 he has successfully defended his PhD thesis and the 5-member examination committee unanimously recommended to the Senate of the University of Cyprus the award of a PhD degree to him. His research interests include computational neuroscience and reinforcement learning with special application on the problem of internal conflict and self-control behaviour. His research has been funded by the Cyprus Research Promotion Foundation and the University of Cyprus.

	Other Presentations Web: https://www.cs.ucy.ac.cy/colloquium/presentations.php
	Colloquia Web: https://www.cs.ucy.ac.cy/colloquium/
	Calendar: http://testing.in.cs.ucy.ac.cy/louispap/XCS-3.0/schedule/cs.ucy.pres.2010.cleanthous.ics

PhD Defense: In Search of Self-Control through Computational Modelling of Internal Conflict, Aristodemos Cleanthous (University of Cyprus, Cyprus), Friday, October 29th, 2010, 9:30-10:30 EET.

The Department of Computer Science at the University of Cyprus cordially invites you to the PhD Defense entitled:

In Search of Self-Control through Computational Modelling of Internal Conflict

Abstract:
This thesis proposes a novel computational model of internal conflict which aims to provide further understanding on this highly complex and perplexing condition of the human brain. In particular, the purpose of this thesis is to identify specific factors which influence and enable internal conflict to be resolved by self-control behaviour. Individuals are likely to experience an internal conflict when evaluating the same outcomes of choice along distinct dimensions or criteria. A value conflict of this sort can be resolved as if it was a result of strategic interaction between rational subagents of the brain. The particular setting for this interaction is a well-studied theoretical game, the Iterated Prisoner’s Dilemma, where the mutual cooperation outcome of the game corresponds to the behaviour of self-control. The computational system developed for the purposes of this thesis realises this particular view of internal conflict by implementing two spiking neural networks as two agents competing in the Iterated Prisoner’s Dilemma, where the agents pursue individual value maximisation through simultaneous but independent learning. This high-level game theoretical approach to the problem of internal conflict incorporates at the same time biological realism through the employed neuronal model, the process of learning, as well as by relating the agents and their actions in the game with particular brain regions and their functioning. In particular, the spiking neural networks comprise of leaky integrate-and-fire neurons, while the learning process is implemented by reinforcement of stochastic synaptic transmission as well as by reward-modulated spike-timing-dependent plasticity with eligibility trace. Moreover, the action of cooperation and defection by each agent maps to a greater relative activity of fronto-parietal and limbic system areas respectively. As demonstrated through numerous simulations, the artificial neuronal system behaved efficiently in the game theoretical framework because the learning agents implemented the optimum result for the system through consistent mutual cooperation. Therefore self-control behaviour can indeed be learned (since it corresponds to mutual cooperation), and as showed by further results, it is enhanced by strong reward-correlated memory. Moreover, the ability of the agents to adopt optimal counter strategies as a response to their competitor’s, enabled the identification of particular value structures that characterise internal conflicts of low and high intensity that promote or hinder the attainment of self-control behaviour. In the process of obtaining the results which are relevant to the problem of self-control behaviour and internal conflict, this thesis work applied for the first time spiking neural agents combined with biological plausible reinforcement learning in a highly demanding multiagent task. In addition, further results with our system showed that high firing irregularity at high rates enhances learning.

Short Bio:
Aristodemos Cleanthous received his B.Sc in Mathematics and Economics at the London School of Economics and Political Science (LSE) in 2002. He further studied at the University College London (UCL), receiving his M.Sc in Computer Science in 2005. Since 2006 he is a PhD candidate at the Department of Computer Science, University of Cyprus. His research interests include computational neuroscience and reinforcement learning with special application on the problem of internal conflict and self-control behaviour. His research has been funded by the Cyprus Research Promotion Foundation and the University of Cyprus.

	Other Presentations Web: https://www.cs.ucy.ac.cy/colloquium/presentations.php
	Colloquia Web: https://www.cs.ucy.ac.cy/colloquium/
	Calendar: http://testing.in.cs.ucy.ac.cy/louispap/XCS-3.0/schedule/cs.ucy.pres.2010.cleanthous.ics