Future CS4 talk: Lin Padgham

linphotoOn Friday 15th March, Professor Lin Padgham, RMIT, Melbourne, will give the CS4 talk “Improving the representation of cognitive agents in agent-based modelling and simulation”

Building 53 Room 4025 2-3pm. Please note this talk is not in the usual Wednesday afternoon slot.

Abstract: 

“Many agent based simulations need to represent humans, their behaviour and their decision making. Although animals can often be modelled well by simple reactive rules based on environment state, this is often inadequate for modelling human behaviours. The Belief Desire Intention (BDI) framework is a well developed theoretical and computational framework that is widely used in developing intelligent agent systems.  Applications are developed by programming interacting agents with goals and plans, to provide a system which does some complex task (e.g. co-ordinating an electronic marketplace or logistics management for a large delivery company), typically in a dynamic environment. The execution engine of such systems then coordinates each agent’s selection of plans and goals, depending on the current situation. We describe how we have integrated a BDI platform with a commonly used ABMS platform (Repast) to take advantage of this more complex, but well structured, modeling of humans available in BDI programming systems/languages.  We also describe a prototype tool to support domain experts and model developers in both understanding and developing the decision making model of an agent using a BDI platform. We also give a brief overview of the RMIT Intelligent Agents group.”

CS4 future talk: Dr Rene Doursat

On Wednesday 6th March, Dr Rene Doursat, Drexel University, Philadelphia, will give theDoursat CS4 talk “Morphogenetic Engineering: the Two-Way Bridges Between Biomodelling, Bioinspired Engineering, and Bioengineering”

Building 53 Room 4025, Highfield Campus, 4-5pm. Refreshments served after the talk.

Abstract:

“Complex systems are large sets of elements that interact locally and produce non-trivial collective behaviours. Natural complex systems, in particular biological, can be a powerful source of inspiration for future technologies. Understanding them by modelling and simulation can help create a new generation of autonomous and adaptive artificial systems, with potential “self‑x” properties mostly absent from traditional engineering. Conversely, closing the loop, these new computing principles can be applied to biomedical challenges, from image processing to “bioware” implementations such as synthetic biology or tissue engineering.

Historically, along these lines, the observation of neurons and genes has given rise to machine learning and evolutionary algorithms. However, these disciplines have also largely shifted their focus to classical optimisation and search problems, away from distributed and emergent computation. In this talk, I want to show new avenues of biomodelling and bioinspired design stressing the importance and benefits of genuine self-organisation in architectured systems, as exemplified by the growth of multicellular organisms. I will present a recent field of research, “morphogenetic engineering” ( http://doursat.free.fr/mebook.html ), which explores the design of complex morphologies that can develop without centralised or external control. Then, I will describe three studies illustrating the possible two-way transfers between modelling and engineering in biological complex systems ( http://doursat.free.fr/devo.html ): (1) a 3D agent-based simulation of early animal embryogenesis drawing from, and compared to, real microscopy imaging data, (2) an artificial evo-devo model of animated organisms in a 3D virtual world, and (3) an attempt at creating a programming language framework to “compile” desired spatial functions or behaviours in a population of bacteria.”

CS4 future talk: Simon McGregor

On Wednesday 24th April, Dr Simon McGregor, will give the CS4 talk “Can Chemicals Think? Learning And Approximate Bayesian Inference In Simulated Reaction Systems”.

Building 53 Room 4025, Highfield Campus, 4-5pm.

Abstract:

“In biology, cognitive processes have been extensively studied in whole organisms, nervous systems, and perhaps to a lesser extent immune systems. The possibility of analogous phenomena at the sub-cellular level has been given relatively little attention. I describe a series of simple experiments showing that potential chemical learning mechanisms can be identified by in silico evolution of simulated reaction systems. In fact, the dynamics can be readily interpreted as an approximate implementation of Bayesian inference. I discuss these results in the context of the complex relationships between information theory, thermodynamics, cognition and life.”

A paper about this research was recently published in PLOS Computational Biology, available for download here.