Julian Togelius

Postdoctoral researcher
Dalle Molle Institute for Artificial Intelligence (IDSIA)
Galleria 2, 6928 Manno-Lugano
Switzerland

Contact:
+41-764-110679(in Switzerland)
+46-705-192088 (in Sweden)
julian@togelius.com

What's new:
I'm setting up a parallel projects webpage at the IDSIA server.

Research interests
After studying philosophy and psychology for some time, and some attempts at participating in the language game that is philosophy of mind, I decided that I was unlikely to contribute to our understanding of the mind through rehashing old philosophers' thoughts. To understand the mind we need to build one. But as no-one knows how to build a mind, we need to find a way to help the mind build itself.

In other words, I'm doing evolutionary robotics, or maybe I should say evolutionary game-playing, or just evolutionary artificial intelligence. I'm using artificial evolution to create neural networks that control the behaviour of (currently simulated) agents, such as robots or racing cars.

Yes, racing cars. I have this idea that computer games are ideal environments in which to develop general intelligence. The main reason is that a well-crafted game (and don't forget that hundreds of man-years go into the development of many commercial games) is a smooth learning task, that starts with relatively modest demands on the player, but in the course of the game requires more and more complex cognitive skills. Moreover, games, with all their sophisticated graphics, provide ample high-dimensional input data for the agent's sensors, something which I think is very important and often overlooked in commonly used robot simulators. And... well, I could go on about this, but I won't - read some of my papers instead!

Of course, I hope my research can improve our understanding of intelligence, evolutionary dynamics, neural dynamics etcetera. I wouldn't mind if I came up with something that could actually be of use in e.g. computer games as well. In any case, there are too many good ideas out there, just waiting to be explored! I recently finished my PhD at the University of Essex, supervised by Simon Lucas. I'm now working as a postdoc at IDSIA, with Juergen Schmidhuber.

Other interests
I'm involved in running the web communities katastrof.nu (swedish) and Galaksen. Very rarely I make music, but when I do something like the ironic, studenticose Swedish rap group Studentz with attitude might emerge. My favourite novelists are Jorge Luis Borges, Umberto Eco and Lars Gustafsson - I guess that says something about me. Oh, and these days I also hava a blog

In the media
Autonomous driving systems aim to drive dirty, New Scientist (23 March 2007)
"Another Step Towards the Driverless Car", Slashdot (23 March 2007)

Organizing activities
WCCI 2008 Simulated Car Racing Competition
CEC 2007 Simulated Car Racing Competition
CIG 2007 Simulated Car Racing Competition

Publications
Julian Togelius, Faustino Gomez and Juergen Schmidhuber (2008): Learning what to ignore: Memetic climbing in topology and weight space. To be presented at CEC.

Julian Togelius, Renzo De Nardi and Alberto Moraglio (2008): Geometric PSO + GP = Particle Swarm Programming. To be presented at CEC.

Alberto Moraglio, Cecilia Di Chio, Julian Togelius, and Riccardo Poli (2008): Geometric Particle Swarm Optimization. Journal of Artificial Evolution and Applications, Volume 2008, Article ID 143624.

Simon M. Lucas and Julian Togelius (2007): Point-to-Point Car Racing; Evolution Versus Temporal Difference Learning. Journal of Information Technology and Intelligent Computing.

Julian Togelius, Simon M. Lucas and Renzo De Nardi (2007): Computational Intelligence in Racing Games. In Norio Baba et al. (eds.) Advanced Intelligent Paradigms in Computer Games

Julian Togelius, Peter Burrow and Simon M. Lucas (2007): Multi-population competitive co-evolution of car racing controllers. Proceedings of CEC.

Alexandros Agapitos, Julian Togelius and Simon M. Lucas (2007): Multiobjective techniques for the use of state in genetic programming applied to simulated car racing. Proceedings of CEC.

Julian Togelius, Renzo De Nardi, Hugo Marques, Richard Newcombe, Simon M. Lucas and Owen Holland (2007): Nonlinear dynamics modelling for controller evolution. Proceedings of Gecco 2007.

Alexandros Agapitos, Julian Togelius and Simon M. Lucas (2007): Evolving controllers for simulated car racing with object-oriented genetic programming. Proceedings of Gecco 2007.

Alberto Moraglio, Julian Togelius and Simon M. Lucas (2007): Geometric Particle Swarm Optimization for the Sudoku Puzzle. Proceedings of Gecco 2007.

Edgar Galvan Lopez, Julian Togelius and Simon M. Lucas (2007): Towards understanding the effects of neutrality on the Sudoku problem. Presented as a poster at Gecco 2007.

Julian Togelius, Renzo De Nardi and Simon M. Lucas (2007): Towards automatic personalised content creation for racing games. To be presented at IEEE CIG 2007.

Simon M. Lucas and Julian Togelius (2007): Point-to-Point Car Racing: an Initial Study of Evolution Versus Temporal Difference Learning. To be presented at IEEE CIG 2007.

Hugo Marques, Julian Togelius, Magdalena Kogutowska, Owen Holland and Simon M. Lucas (2007): Sensorless but not Senseless: Prediction in Evolutionary Car Racing. To be presented at IEEE ALife 2007.

Julian Togelius, Renzo De Nardi and Simon M. Lucas (2006): Making racing fun through player modeling and track evolution. Proceedings of the SAB Workshop on Adaptive Approaches to Optimizing Player Satisfaction.

Julian Togelius and Simon M. Lucas (2006): Arms races and car races. Proceedings of Parallel Problem Solving from Nature.

Julian Togelius and Simon M. Lucas (2006): Evolving robust and specialized car racing skills. Proceedings of the IEEE Congress on Evolutionary Computation.

Renzo De Nardi, Julian Togelius, Owen Holland and Simon M. Lucas (2006): Evolution of Neural Networks for Helicopter Control: Why Modularity Matters. Proceedings of the IEEE Congress on Evolutionary Computation.

Alberto Moraglio, Julian Togelius and Simon M. Lucas (2006): Product Geometric Crossover and the Sudoku Puzzle. Proceedings of the IEEE Congress on Evolutionary Computation.

Julian Togelius and Simon M. Lucas (2005): Evolving Controllers for Simulated Car Racing. Proceedings of IEEE Congress on Evolutionary Computation, 1906-1913; winner of the best student paper award!

Julian Togelius and Simon M. Lucas (2005): Forcing neurocontrollers to exploit sensory symmetry through hard-wired modularity in the game of Cellz. Proceedings of the IEEE Symposium on Computational Intelligence and Games, 37-43.

Julian Togelius (2004): Evolution of a Subsumption Architecture Neurocontroller. Journal of Intelligent & Fuzzy Systems 15:1, pages 15-20.

Dissertations
Julian Togelius (2007): Optimization, Imitation and Innovation: Computational Intelligence and Games. PhD Thesis, Department of Computing and Electronic Systems, University of Essex, UK.

Julian Togelius (2003): Evolution of the layers in a subsumption architecture robot controller. MSc dissertation, COGS, University of Sussex, UK.

Julian Togelius (2001): P-medvetande och andra medvetanden (P-consciousness and other minds). Bachelor's dissertation, Theoretical Philosophy, Lund University, Sweden.

CV
In case you're interested, please have a look.

Source code
Java code for the experiments above is available here.