Directed Information Measures in Neuroscience

I'm really please to announce the publication of our new edited book:

"Directed Information Measures in Neuroscience"
edited by Michael Wibral, Raul Vicente, Joseph T. Lizier
in series "Understanding Complex Systems",
Springer, Berlin, 2014.

About -- Downloads (via Springer) -- Purchase (via amazon)

The book  grew out of a workshop I co-organised with Michael and Raul in Frankfurt in April 2014 -- NeFF-Workshop on Non-linear and model-free Interdependence Measures in Neuroscience. Our workshop focussed on the use of transfer entropy in computational neuroscience. We managed to attract several good speakers from this field, including Daniele Marinazzo, Daniel Chicaharro, Luca Faes and Vasily Vakorin, as well as a good crowd of participants, many of whom were quite knowledgeable in this field, such as Demian Battaglia. In our humble opinion, the meeting was quite a success, culminating in lively discussion sessions at the end of each day. We were delighted to host Leontina Di Cecco from Springer at the workshop, and the book project grew from our discussions there.

The contributions were mainly decided at the workshop, and with chapters contributed from the aforementioned authors, we managed to span most of the research taking place on information transfer in neuroscience. The book serves as a thorough introduction to measuring directed information transfer in computational neuroscience, how this is being applied and what it can reveal, and what directions this research may take in the future. We're really happy with the end result, and even more pleased that we can share it with you now.

For a little more information, here's the teaser from the back of the book:

Analysis of information transfer has found rapid adoption in neuroscience, where a highly dynamic transfer of information continuously runs on top of the brain's slowly-changing anatomical connectivity. Measuring such transfer is crucial to understanding how flexible information routing and processing give rise to higher cognitive function. Directed Information Measures in Neuroscience reviews recent developments of concepts and tools for measuring information transfer, their application to neurophysiological recordings and analysis of interactions. Written by the most active researchers in the field the book discusses the state of the art, future prospects and challenges on the way to an efficient assessment of neuronal information transfer. Highlights include the theoretical quantification and practical estimation of information transfer, description of transfer locally in space and time, multivariate directed measures, information decomposition among a set of stimulus/responses variables and the relation between interventional and observational causality. Applications to neural data sets and pointers to open source software highlight the usefulness of these measures in experimental neuroscience. With state-of-the-art mathematical developments, computational techniques and applications to real data sets, this book will be of benefit to all graduate students and researchers interested in detecting and understanding the information transfer between components of complex systems.

Information theory: questions and answers

Information theory is fundamentally about questions and answers.

We understand information itself in terms of questions and answers: 1 bit of information is the uncertainty in the answer to a question with a 50-50 outcome, e.g. "will this coin flip give tails?".

Just as importantly though, the measures of information theory themselves are all about questions and answers too.

For the basic measures, the questions they ask seem fairly obvious. The Shannon Entropy asks:

"How much uncertainty is there in the state of this variable X?".

Mutual information asks "how much information does the state of variable X tell me about the state of Y?", while conditional mutual information asks "how much information does the state of variable X tell me about the state of Y, given that I already know the state of Z?"

But I want to make a few more subtle points about these questions and answers.

In my opinion (which is of course the only correct one), the answers that the measures give are always correct. If you think they're wrong, then you're asking the wrong question, or have malformed the question in some way. There are plenty of ways to do this, or at least to inadvertently change the question that you're asking.

I see the sample data itself as part of the question that a measure is answering. When you estimate the probability distribution functions (PDFs) empirically from a given sample data set, your original question about entropy really becomes:

"How much uncertainty is there in the state of this variable X, given what we're assuming to be a representative sample of realisations x of X here?"

Of course, your representative sample could simply be too short, and thereby completely misrepresent the PDF. Or you could get into trouble with stationarity (1) of the process - you might implicitly have appended "given what we're assuming to be a representative stationary sample here" to the question, but that assumption may not be true.
In both cases, the measure will give the correct answer to your question, but it might not be the question you really intended to ask.

As another way of inadvertently changing the question, one must realise that for the same information-theoretic measure, different estimators (or indeed different parameter settings for the same estimator) answer different questions. Take the mutual information, for example, which one could measure on continuous-valued data via (box) kernel estimation. Using this estimator, the measure asks: "how much information does knowing the state of variable X within radius r tell me about the state of variable Y within radius r?" Clearly, using different parameter values for r amount to asking different questions - potentially the questions are very different if one uses radically different scales for r. Going further, one could measure the mutual information using the enhanced Kraskov-Grassberger kernel estimation technique. With this estimator, the mutual information measure asks "how much information does knowing the state of variable X tell me about the state of variable Y, to the precision defined in their k closest neighbours of the sample data set in the joint X-Y space?" Apart from that being something of a mouthful, it's obviously a different question to what the box kernel estimation is asking. And again, changing the parameter k changes the question being asked as well.

So to reiterate, information theory is fundamentally about questions and answers - the better you can keep that in mind, the better you will understand information theory and its tools.


UPDATE- 13/12/12 - My colleague Oliver Obst provided a perfect quote about this: "Better a rough answer to the right question than an exact answer to the wrong question" - attributed to Lord Kelvin.

-------------------
Footnotes:
(1) Here's a controversial statement: I suggest that it can be valid to make information-theoretic measurements on non-stationary processes. This simply changes the question that is being asked to something like: "how much uncertainty is there in the state of this non-stationary variable X, if we don't know how the joint probability distribution of the non-stationary process is operating at this specific time, given what we're assuming to be a representative sample of the joint probability distribution weighted over all possible ways it may operate?". Now, obviously that's quite a mouthful, but I'm trying to capture that intuition that one could validly consider how much information it takes to predict X if we don't know the specifics of the non-stationarity at this particular point in time, but do know the overall distribution of X (covering all possible behaviours). So long as one bears in mind that a different question is being asked (indeed a question that is quite different to the intended use of the measure), then certainly the answer can be validly interpreted. Of course, the bigger issue is in properly sampling the PDF of X over all possible behaviours, but that's another story.

Workshop on Non-linear Interdependence Measures in Neuroscience

I'm pleased to announce the NeFF-Workshop on Non-linear and model-free Interdependence Measures in Neuroscience and TRENTOOL course which will be held at Goethe University Frankfurt, Germany on April 26-27, 2012 (hosted by the MEG Unit of the Brain Imaging Centre, Frankfurt).

The synopsis from the workshop announcement is as follows:

Understanding complex systems composed of many interacting units, such as neural networks, means understanding their directed and causal interactions. If the units in question interact in a nonlinear way, as it can be assumed in neural networks, we are faced with the problem that the analysis of interactions must be blind to the type of interaction if we want to cover all possible interactions in the network, as we may not know the type of nonlinear interaction a priori. Prematurely limiting our search to specific models, nonlinearities or, even worse, linear interactions may block the road to discovery. Novel model-free techniques for the quantification of directed interactions from information theory offer a promising alternative to more traditional methods in the field of interaction analyses, but also come with their own specific challenges. This symposium brings together the most active researchers in the field to discuss the state of the art, future prospects and challenges on the way to an model-free, information theoretic assessment of neuronal directed interactions.

I'm happy to be co-organising this workshop with Michael Wibral (head of the MEG Unit, Brain Imaging Center, Goethe University Frankfurt) and Raul Vicente  (Frankfurt Institute for Advanced Studies).

We've got several speakers lined up to talk about their work in this field, particularly using information-theoretic tools including the transfer entropy. The speakers include some collaborators of mine (e.g. Mikhail Prokopenko, Paul Williams), many others I'm looking forward to meeting (e.g. Stefano Panzeri, Luca Faes), and the organisers of course :).

Plus there will also be a workshop on Michael and Raul's Transfer Entropy toolbox (TRENTOOL), which is designed to provide effective network analysis on neuro data sets in Matlab. I'm looking forward to playing around with this more myself, I've already got a project and some data in mind.

We're hoping to get lots of participants (though space is limited) - full details on how to register are available at the workshop website - http://www.neff-ffm.de/de/veranstaltungen/seminars/workshop.php

I hope to see you there!

Identifying influential spreaders and efficiently estimating infection numbers in epidemic models: a walk counting approach

Updated 26 October - our paper has been published in Europhysics Letters 99 68007 (2012) doi:10.1209/0295-5075/99/68007 - we've also updated the preprint on arXiv with the revised material.

Update 3 December - MPI made a press release about the paper, and New Scientist (German edition) published an article about it (in German).

-------------------------------------------------------------------

The news at this end is that Frank Bauer and I just submitted a new preprint on arXiv:

F. Bauer and J.T. Lizier, "Identifying influential spreaders and efficiently estimating infection numbers in epidemic models: a walk counting approach", MPI MIS Preprint 1/2012, arXiv:1203.0502, 2012.

We introduce a new method to efficiently approximate the number of infections resulting from a given initially-infected node in a network of susceptible individuals. Our approach is based on counting the number of possible infection walks of various lengths to each other node in the network. We analytically study the properties of our method, in particular demonstrating different forms for SIS and SIR disease spreading (e.g. under the SIR model our method counts self-avoiding walks). In comparison to existing methods to infer the spreading efficiency of different nodes in the network (based on degree, k-shell decomposition analysis and different centrality measures), our method directly considers the spreading process and, as such, is unique in providing estimation of actual numbers of infections. Crucially, in simulating infections on various real-world networks with the SIR model, we show that our walks-based method improves the inference of effectiveness of nodes over a wide range of infection rates compared to existing methods. We also analyse the trade-off between estimate accuracy and computational cost, showing that the better accuracy here can still be obtained at a comparable computational cost to other methods.

Epidemic spreading in biological, social, and technological networks has recently attracted much attention. The structure of such networks is generally complex and heterogeneous, so a key question in this domain is: "Given a first infected individual of the network (patient zero) - how likely is it that a substantial part of the network will become infected?" It is, of course, of particular interest to identify the most influential spreaders. This knowledge could, for instance, be used to prioritise vaccinations.

The most obvious and direct way to address this question is to estimate the number of infections by running simulations of the infection model. There are several well-known infection models which can be used to simulate diseases with different properties. These include the SIR (susceptible-infected-removed) model for diseases where a subject may only be infected once (due to either recovery with full immunity or death), and the SIS (susceptible-infected-susceptible) model for diseases where infected subjects recover and become susceptible to reinfection. To run a simulation, one must consider the network structure connecting susceptible individuals, and the infection rate or probability β that an infected individual will infect a given neighbour.

One problem with running simulations however, is that it takes a lot of computational time to obtain appropriate accuracy. For example, for an SIR model we run on a network of 27519 nodes and 116181 undirected edges, 10000 simulated initial infections per node for around 20 values of infection rate β took 2000 hours to simulate. Of course, this runtime does not scale well with the size of the network, while for real-world problems it is generally the large networks that we are genuinely interested in.

As such, it would be useful to find a more efficient way than full simulation to estimate infection numbers, and/or to be able to use local network properties of nodes to understand their spreading efficiencies.
So the problem that we are addressing here is two-fold:

1. How to efficiently estimate the number of infections resulting from a given initially-infected node in a network of susceptible individuals?
2. What network structural properties which are local to the initially-infected node are most useful for predicting how well disease will spread from it?

In fact, there has been a lot of work recently trying to find local network properties of nodes that are useful in predicting the relative spreading influence of different initially-infected nodes in a network. This attempts to address problem 1, but additionally gives very useful insight into how local network structure can promote or inhibit disease spreading (i.e. problem 2). The properties other authors have investigated range from simply examining out-degree, to k-shell analysis, to various measures of node centrality in the network (e.g. eigenvector centrality). And the good news is that you get a surprisingly accurate insight into the relative spreading efficiency of the various nodes with these very simple measures. Such work has attracted a lot of attention, for example being published in Nature Physics.

However, we observed two issues with these approaches:

1. They only infer the relative spreading efficiency of initially infected nodes; i.e. they do not provide an estimate of the actual numbers of infections resulting from each node. These actual infection numbers could be very important in many applications.
2. While the existing inference measures do a good job, they do not actually directly consider the mechanics of disease spreading. As such, there is still room for improvement. As an example of potential improvement areas: none of the above-mentioned measures change their relative inference with the rate of infection β.

So, we sat down and thought hard about the local network properties that best relate to the mechanics of disease spreading. We focussed on the fact that disease spreads on a network in the manner of a walk. The disease can only reach node B from node A on a walk from A to B, where every node on that walk is also infected. Our basic premise is that the count of the number of walks from the initially infected node to other susceptible nodes (an approach known as walk counting) should be a local network structural property that gives good insight into disease spreading. The idea had been previously raised in the literature, but not properly examined.

We developed the idea further, working out the mathematics to turn these walk counts into estimates of infection numbers, as a function of infection rate β. Interestingly, different types of walks are involved for different disease spreading models; e.g. for SIR spreading one is only interested in self-avoiding walks (since no node can be infected twice), whereas for SIS spreading one is interested in any type of walk. Our estimates are not perfect, and we identify where and how known errors will occur. However, the estimates have several very important and useful properties compared to other approaches:

1. Our technique provides estimates of actual numbers of infections from a given initially-infected node, which is more useful than inferred relative spreading efficiency alone. It's useful to know who the most influential spreaders are, but you also want to know the extent to which they will spread the disease.
2. Importantly, testing with simulations on various social network structures reveals that our technique infers more accurate relative spreading efficiencies than those of the aforementioned previously published techniques over a wide range of infection rates β (up to the lower super-critical spreading regime). This is because our technique directly considers the mechanics of disease spreading.
3. And our technique has excellent computational efficiency. Note that there is a trade-off in our technique between computational efficiency and higher accuracy: by considering only short infection walks, our algorithm runs faster, but better accuracy is generally obtained by considering longer walks. Our short-walk estimates can be made in approximately the same run-time as the aforementioned techniques, but with greater accuracy in inferring relative spreading efficiency. Our longer-walk estimates produce better accuracy again, and do so with orders of magnitude less runtime than simulations of the disease spreading process which obtain the same accuracy.

As such, we show that our walk counting approach provides the following unique combination of features: they are the most relevant local network structural feature to infer relative disease spreading efficiency, provide estimates of actual infection numbers, and are computationally efficient.

Comments/suggestions welcome, of course ...

10th International Conference on Cellular Automata for Research and Industry (ACRI 2011)

The first call for papers is out for the 10th International Conference on Cellular Automata for Research and Industry (ACRI 2012), to be held on Santorini Island, Greece, September 24-27, 2012.

The main website is at http://acri2012.duth.gr/
Important dates at http://acri2012.duth.gr/dates.html - submissions are due March 19, 2012.
Other important information includes that the proceedings will be published in Springer LNCS.

This year I am part of the program committee, which should be quite interesting. I've come across many papers from this conference series, including "Local Information in One-Dimensional Cellular Automata" which influenced my own work on filtering CAs. I'm looking forward to going, and hope to see you there.

Topical Issue on Guided Self-Organization

Following the success of the third and fourth Guided Self-Organization workshops (this year and last), there is a call for papers out for a topical issue on Guided Self-Organization in Advances in Complex Systems next year.

Important dates are:

• expression of interest (tentative title and list of authors) to guest editors : 4. November 2011
• submission to ACS: 31 January 2012
• notification: 30 April 2012
• camera-ready papers: 31 May 2012

Full details of the CFP are at http://informatics.indiana.edu/larryy/gso4/cfp/index.html and an excerpt is below:

The goal of Guided Self-Organization (GSO) research is to leverage the strengths of self-organization while still being able to direct the outcome of the self-organizing process. The ACS Topical Issue on Guided Self-Organization aims to condense the current state-of-art in guided self-organizing systems, including, but not limited to information- and graph-theoretic foundations of GSO and the information dynamics of cognitive systems.

A number of attempts have been made to formalize aspects of GSO within information theory and dynamical systems: empowerment, information-driven evolution, robust overdesign, reinforcement-driven homeokinesis, predictive information-based homeokinesis, interactive learning, etc. However, the lack of a broadly applicable mathematical framework across multiple scales and contexts leaves GSO methodology incomplete. Devising such a framework and identifying common principles of guidance are the main themes of GSO.

Papers need not be regarding work presented at the workshops, new work is also solicited. Good luck with your submissions!

GSO 4

A quick note to promote the Fourth International Workshop on Guided Self-Organization (GSO 4):

The goal of Guided Self-Organization (GSO) is to leverage the strengths of self-organization while still being able to direct the outcome of the self-organizing process. The GSO-2011 workshop will bring together invited experts and researchers in self-organizing systems, with particular emphasis on the information- and graph-theoretic foundations of GSO and the information dynamics of cognitive systems.
...
The following topics are of special interest: information-theoretic measures of complexity, graph-theoretic metrics of networks, information-driven self-organization (IDSO), applications of GSO to systems biology, computational neuroscience, cooperative and modular robotics, sensor networks, and cognitive modeling.

Some good friends of mine have been behind this series (this year, Daniel Polani, Larry Yaeger, and my old supervisor Mikhail Prokopenko). The series started at our lab in Sydney 3 years ago, and it's pleasing to see that it has really got some momentum behind it now.

Unfortunately I have to miss it this year, but if this sounds like your field then I recommend that you go, as this will be an excellent meeting.

Abstracts are due by July 31, the workshop itself is on Sept 8-10 2011 in Hertfordshire, UK.

Carbon tax shenanigans

It's been a little strange watching all the debate at home about the incoming carbon tax. The commentators on the right are getting so frothy-mouthed and vicious about the whole thing. Hardly a surprise I suppose.

The weirdest thing is the loss of perspective. I think it's best summarised in this blog post (tip to Elliot):
http://www.heathenscripture.com/you-shut-your-goddamn-carbon-taxin-mouth/
It's a great shot of perspective there. And I have to agree - if you can't afford $10 a week out of your $100k+ income for something for your kids' futures like this, my heart bleeds. Really.

I also wanted to share something I saw on the BBC news this morning. After a story on the impending (real) Italian financial crisis, they reported that consumer confidence in Australia had reached a low point, noting with unhidden incredulity that this was despite (and I paraphrase) "near zero unemployment, strong growth and record standard of living" but seemed "related to fears about a carbon tax". It's difficult not to feel embarrassed about that.

I must start writing again ...

Well it's been a while, and a lot has happened since I last wrote.

In the last 18 months or so, I've submitted my PhD thesis, worked some more hours in my software engineering job, wrote up a few papers, graduated, moved to Leipzig, Germany and started as a postdoc at the Max Planck Institute for Mathematics in the Sciences.

Life hasn't exactly settled down, but I am planning on writing about all of the above in the near future ...

Tweeting

So I've been seduced by micro-blogging lately instead of doing much writing over here. It makes me think of the trend away from test match cricket towards the twenty-20 variety ...

Anyway. I have to say that I've been enjoying getting into it. Follow me over at twitter: @jlizier

Writing up

Well, that came around much more quickly than I was expecting. The last three and a half years doing my PhD have been really enjoyable, so on one hand writing up my thesis makes me a little sad to be finishing up. On the other hand, it's quite exciting and I'm looking forward to seeing the bound, finished product.

Anyway, I'm well on the way at the moment, with about 4.5 chapters out of the 6 technical chapters done. And I'm only just hitting the writer's block stage; well that's an overstatement, I'm just having a little trouble getting the right tone on this chapter. Nothing serious. I'm still on track for a pre-Christmas submission. As I said to my supervisors, the only risk is that they won't provide the feedback fast enough :).

I'm looking forward to a nice holiday over Christmas and January, then who knows where I'll end up next year. I put my first post-doc application in last week, I'm working on another one, and planning to put in a local application early next year. Watch this space.

Neural Computation discussion group

Today I'm just going to run a quick advertisement for a new series we're starting in our discussion group on"Neural Computation". It's on Fridays @ 2 pm at CSIRO Marsfield (north-west Sydney), but is open to anyone from outside who is interested.

More details, including the schedule of talks, is available at http://www.prokopenko.net/entropy.html

As it happens, stumbling upon a link for these talks and subsequently coming along to them is how I met Mikhail which led to me starting my PhD. So if you're interested in the area, do come along as you never know where it might lead you!

Funny stunt

Here I was thinking that I was procrastinating about writing up my PhD. Not as much as these guys:
http://dataphiles.blogspot.com/2009/09/machine-learning-protest-at-g20.html
One of the funniest things I've seen for a while ...

Apology to Turing

Not mine, the British government's.

Better late than never.

Information, computation and complex systems workshop

A quick post to advertise the Information, Computation and Complex Systems satellite meeting, which is part of ECCS 09.

From the workshop page:

This workshop will bring together mathematicians and scientists to discuss methods and applications of information theory to complex systems. This includes the perspective of complex systems as computers.
The question of good measures of complexity alone is vital for complex systems research. Many proposals have been made, quite a few based on information theory. The workshop addresses the following topics:

1. Mathematics of information theoretic tools for complex systems
2. Information theory applied to complex systems
3. Complex systems as information processors

The workshop will provide a forum for discussing the various existing information-theoretic and computation-theoretic tools, and their use in complex systems.
The focus will be on the mathematics of information and computation theory applied to complex systems. We encourage both theorists and experimentalists to attend, with either an information theoretic approach or a need for an information theoretic approach.

I think this will be a really interesting workshop, as it directly relates to the focus of my PhD. The invited presentations should be quite good. I'm hoping to get the opportunity to present my own approach of the local information dynamics of distributed computation in complex systems, and hope to meet some people there with the need for this kind of approach.

Submissions are apparently open until Mon 24/8/09, hope to see you there.

Update 1/9/09 - my submission "Coherent local information dynamics in complex computation" was accepted, so I'm now looking forward to presenting at ICCS. The schedule has been posted.

The Ashes 2009

Tonight (Australian time) marks the start of the 2009 Ashes series in the UK. For the uninitiated, The Ashes is the "trophy" for Australia vs England cricket series. These series occur roughly every two years, with a series hosted by each country each four years. For an Australian player or fan, an Ashes series in England is as good as it gets.

To say I am looking forward to the coverage is an understatement. I would really love to get to England for the series one day. I'll be nearby in Germany during the last game of the series this year, but unfortunately all of the tickets are sold out.

As I'm sitting here watching the first few overs, I've been thinking about how I can almost mark progress in my life by what I was doing while watching the Ashes from England every four years. In 1989 I was just finishing primary school. I hadn't been interested in or playing cricket for long, but it sure took up a lot of my thinking. In 1993 I was mid-way through high school, and by this point was taking my studies fairly seriously as I had realised what I could achieve academically. I remember watching the games up late after coming home some of the first times I was allowed to go out with my friends at night. By 1997 I was a couple of years into uni, really enjoying engineering. I was also going out with my wife (then girlfriend) by then and my enduring memory of that series is watching it at her place after everyone else was asleep. 2001 was a big year for growing up: first full-time job, moving out with my girlfriend and thinking about getting married. It was nice watching that series late at night in a place I could call my own. Much better was 2005, when I could watch it in an apartment I owned rather than rented. By this time, we were married, and I had established myself at work though I was thinking a lot about how to go about doing a PhD as I was watching the games. This time, I'm sitting in our own house, am writing up my thesis, and my wife has begun her PhD studies. This time around I'll be occupied by what I/we are going to do next year once I'm finished. I wonder where I will be in four years time?

The information dynamics of cascading failures in energy networks

Disclaimer: shameless self-promotion follows.

So our submission:

Joseph T. Lizier, Mikhail Prokopenko, David J. Cornforth, "The information dynamics of cascading failures in energy networks"

to ECCS 2009 was accepted, and I'll be presenting it in the Policy, Planning and Infrastructure track currently on the Friday morning of the conference.

The abstract is as follows:

Small failures in electrical energy networks can lead to cascading failures that cause large and sustained power blackouts. These can disrupt important services and cost millions of dollars. It is important to understand these events so that they may be avoided. We use an existing model for cascading failures to study the information dynamics in these events, where the network is collectively computing a new stable distribution of flows. In particular, information transfer and storage across the network are shown to exhibit sensitivity to reduced network capacity earlier than network efficiency does, and so could be a useful indicator of critical loading. We also show that the local information dynamics at each node reveals interesting relationships between local topological features and computational traits. Finally, we demonstrate a peak in local information transfer in time coinciding with the height of the cascade's spread.

In a nutshell, this paper describes an application of our framework for the information dynamics of distributed computation to the phenomena of cascading failures on networks. The focus is on energy networks, though the results are applicable to other types of networks, e.g. transport.

Information dynamics may at first not seem applicably to cascading failures, but there are a few good reasons for the application here. First, cascading failures are akin to damage spreading phenomena, and both are often cited as mechanisms of information transfer in networks: it is useful to explore this quantitatively. Further, when a cascading failure occurs, the network is actually computing a new stable state (or attractor), so quantifying the information dynamics is a direct study of this computation. To underline all that, I really like this quote from Melanie Mitchell's new book:

The phenomena of cascading failures emphasizes the need to understand information spreading and how it is affected by network structure.

Primarily, the results show that we get maximisations of information transfer and storage in the network near the critical phase, aligning with our findings in Random Boolean Networks (RBNs) in a paper at ALifeXI last year. We also find some interesting relationships between topological properties of the individual nodes and their own local information dynamics.

From here, I'll be combining this work with that on RBNs in my PhD thesis, and probably seeking to make a journal submission from their combination.

Second International Workshop on Guided Self-Organisation (GSO-2009)

Just a quick post to advertise The Second International Workshop on Guided Self-Organisation (GSO-2009) which I'm planning on attending in August.

A few pertinent clips from the workshop's website:

"... by its very nature, self-organization more often than not has its own way. To be useful in practice, methods of guiding self-organization towards prespecified goals have to be developed. Adding and controlling constraints provides one possibility to this end.
Many properties of self-organisation can be characterised formally (e.g., information-theoretically). However, the lack of agreement of what is meant by complexity, constraints, etc, as well as a common methodology across multiple scales leaves any definition of self-organisation somehow vague, indicating a clear gap. Filling this gap and identifying common principles of guidance are the main themes of GSO-2009. The workshop will put particular emphasis on principles based on information flows through the perception-action loop of embodied systems."

GSO-2008 was a really interesting week, so am hoping for the same level of inspiration this time round.

Registration is open, hope to see you there.

Directed information structure

The good news is that we recently received a notice of acceptance of an abstract we submitted to CNS*2009 (July 18-23) in collaboration with John-Dylan Haynes and Jakob Heinzle from the Bernstein Centre for Computational Neuroscience (BCCN) in Berlin.

Titled "Directed information structure in inter-regional cortical interactions in a visuomotor tracking task", it deals with a method for identifying directed information structure between distinct regions (of variables) in a large multi-variate set. The method identifies an interesting hierarchical structure for the given visuomotor task.

My supervisor will be at CNS to present the poster, and the abstract will be included in a supplement to BMC Neuroscience soon. We're currently working on a more complete journal paper reporting on this experiment.

CNS*2009 should be a good meeting, particularly the Methods of Information Theory in Computational Neuroscience workshop. Wish I was going!

ECAL and ECCS calls for papers

Just thought I'd share two calls for papers for conferences that I will be / have submitted to this year.

The European Conference on Complex Systems (ECCS09) will be held from 21-25 September 2009, at the University of Warwick, UK. I've never been to ECCS before, but I've heard good things from several people about it. I like the way it appears to be a real melting pot of all areas of complex systems science, so I'm looking forward to seeing some interesting perspectives there. On that note, I think there will be some interesting applications related papers there, e.g. in the Policy, Planning and Infrastructure track. I also like the tiered submission structure, where you can submit 2, 6 or 15 page papers, and (if accepted) get a poster, 20 min or 40 min presentation: it gives you choice, and appropriate relative reward for work. We've submitted a paper on cascading failures in energy networks (more details if we're accepted). The first submission deadline has passed, though they have two more deadlines coming up (19/4 and 3/5): apparently slots will be filled on a "first arrival - frist serve policy".

The 10th European Conference on Artificial Life (ECAL2009) will be held from September 13-16 2009 in Budapest, Hungary. I was at the last ECAL in Lisbon in 2007 and thoroughly enjoyed it. It's a good crowd, with a nice mix of biologists and computer scientists. We're currently working on a paper combining some of my work on information dynamics with my PhD colleague Mahendra Piraveenan's work on network topological measures: this is something I had wanted to do for a while, but isn't in the form I thought it would be (nothing wrong with that though). More details if we're accepted. Anyway, ECAL are now following the lead of ALifeXI in allowing abstract only submissions (which I think is fine in principle), and allowing both to have presentation slots (this I'm not sure about - I had the impression that some, not all, of what came through the abstract only channel was under-prepared; I prefer the ECCS approach). Paper submission is by April 30.

Hope to see you there! (assuming we get accepted...)