Category Archive

Before I move on to my daily research routine I will first attempt to reflect on the conference. Hopefully this way I can reduce the chances of forgetting it ever happened. I won’t go into much depth but only some of the things I remember most clearly.

ECAL felt like an extended-family-of-the-CCNR week-long meeting. And I think that’s a good thing. The most exciting organisational differences between this ECAL and the previous one, as well as at least the two last Alifes, had to be the panel discussions at the end of each day. Although the discussions were not always insightful (see point towards the end on information theory and dynamical systems), they usually reflected the state of something that’s happening in the Alife community. Another good thing about this conference was the stronger congruence throughout each day’s theme as embodied by the keynotes, talks and posters presented in each.

There were several talks that I particularly enjoyed. Randall Beer’s and Janet Wile’s keynotes were the most exciting for me. I liked them the most because of the clearness of their motivations, the relevance of their scientific questions, the concreteness of the frameworks they propose, and the in-depth and systematic approach that they usually take, among other things. What I particularly liked about Randy’s was his outlook for future directions of research. I also found inspiring his emphasis on the idea that the use of dynamical systems analysis was not impossible to achieve, but on the contrary quite tractable, as long - of course - as education on the language and the tools were available to the community. I very much liked how Janet Wiles was so aware of her role as a supervisor of grad students. She kept bringing the topic of ‘approaches to grad students’ up any chance she had, unfortunately I felt most of the times it was glossed over by the rest and usually put to one side.

There were other good talks. I enjoyed Ezequiel’s keynote because of its exploratory nature and the effort to tie the sort of work that evolutionary robotics people do (and not very unlike mine) together with important philosophical strands. Even though Jun Tani and Takashi Ikegami didn’t give keynotes, from their interventions a similar (if not more exploratory still) approach was always present. This was also refreshing. Particularly, because they were so honest about mostly doing it because it was fun for them. Interestingly enough, although I didn’t get most of Brian Goodwin’s points, his way of addressing us was very inspiring too.

Although I’m not usually amused by engineering feats, I was genuinely impressed with the work of Rudolf Bannasch. Undoubtedly, I was better able to appreciate the extent of his contributions precisely because he made it so clear throughout his talk that his motivation was purely in engineering. This is not typical in Alife. What is more characteristic is that certain disguised-in-science-but-engineering-at-heart works leaves a lot of people confused and makes me in particular obsess throughout the duration of their exposition about which scientific question is the one we are tackling. Also, one usually hears about how understanding of biological organisms is going to impact how we as engineers design machines, but it is not everyday that you see biologically-inspired machines out there in the world. His talk made it seem otherwise which was great. In some respects, I guess Dario Floreano’s talk also helped give that feeling a little bit.

As usual, it is the workshops that stand out the most for me. Unfortunately they all happen in parallel on the same day, so you can only attend to one or two. I attended Rachel Wood’s (and others) ‘dynamics of development’ workshop. Among the participants were Linda Smith, Jane Wiles, Jun Tani, Luc Bertouhze, Ezequiel Di Paolo. Lots of time for discussions and comments. I was fortunate to have been given the chance to comment on Josh Bongard’s latest internal self-modeling robot work and how it might go beyond the metaphor to answer questions relevant to developmental psychologists. I took the opportunity to refresh skepticism about the usefulness of sense-model-plan-act approaches in Robotics when they are used as scientific tools. Overall, I was mostly surprised with the readiness of a good proportion of the participants - in particular people whose work is very much along the lines of a situated, embodied and dynamical systems perspective - to take for granted that we have (or eventually develop) internal models of ourselves that we can ‘use’ to simulate things in.

Worth echoing here are two points that Randy brought up during this workshop. The first is one we are all very familiar with. Basically being very aware of not confusing engineering and science. Particularly of not falling into the trap of going to the scientific meetings and talking about the engineering and viceversa. The second point worth mentioning is regarding the relevance of physical or simulated models. Although I was familiar with the discussion, I hadn’t seen it from the perspective that Randy pointed out until then. His point was that the use of physical robots need not be, by default, more assumptions free nor more realistic, nor better in any a priori way than simulations. He gave a good example of how sometimes simulations can be more biologically realistic than physical models, and how this depended on the proof of how the relevant biomechanical properties compared to the real organism being modeled and not simply on the physicality of the robot - as choices of materials, dimensions, etc. could be very much inappropriate. This is how I interpreted and recall those two points now anyways.

I also took part in the ‘dynamics of learning and neuromodulation’ workshop, of course. That went really really well. Lots of people turned up. In fact some people told me it was one of the busiest workshops there. There was lots of discussion too. Some of it was unfortunately spent clarifying the two different levels of defining learning: mechanistic and behavioral. I understand that it is something that takes a while to assimilate, but I would have thought by now (and in the context of that workshop) that everybody would have been aware of the difference. Some people would inevitable (no matter how many times we reminded them) kept slipping back into a ‘learning as mechanism’ perspective. When that wasn’t the case however, an interesting discussion of what was behavioral learning dominated the discussion - with particularly relevant insights coming from Randy, Anil Seth, Chrisantha and Seth Bullock. This was particularly relevant for me as most of the issues that were discussed are very relevant for my thesis.

There were many good talks and posters as well. But it will take me longer to realise which are relevant for me and which aren’t.

Lisboa was great of course. I’m sure I don’t need to say much there. Undoubtedly, Barrio Alto served as the secondary venue for the conference.

One of the panel discussions opened up a debate about information theory and dynamical systems theory which was unfortunately not too insightful (it has been followed in the conference blog). For two reasons mainly. First, most people seemed to equate dynamical systems theory with particular dynamical neural models and information theory with computation. This obviously leads to people talking across each other about a lot of senseless stuff. Several people pointed this out during the discussion but a few minutes later the discussion would slip into loaded notions of the terms. Second, because people would sometimes argue that one was ‘a better tool’ than the other. This doesn’t make sense, on his own anyway. Better at what? Both of them are tools that can be used to analyse dynamical (and non-dynamical) agent/environment systems. What I think some people may have meant is that there can be certain types of question that one tool will be able to answer better than the other (and of course, some questions which might be better answered by using both or neither tools). Dynamicists, as they are sometimes called, think that certain aspects of interactions (such as relative timing, morphologies, biomechanics, etc.) are important for what they understand as cognition. Therefore, dynamical systems tools are most of the times the more appropriate tool to address the sorts of questions that they have about their systems of interest. On the other hand, informationists (or however these people are called), have very different types of questions about their systems. This is sometimes because of the nature of the systems which they study (such as throwing a dice with clearly labeled sides and discrete events in time), but ultimately the type of system under study should not force one to study it with one tool or the other. It is their type of questions that are different. For example, they might be interested in how much mutual information an agent shares with its environment. It is questions of this nature that they think are relevant for their understanding of cognition, for which information theory is a mighty powerful tool. In summary, I don’t think there is a debate worth having about which is a better tool. Particularly, in the absence of the context of which question and which system the user of the tool is trying to answer. It is more interesting to debate which questions we are asking about our agent/environment systems and the tools will simply follow.

Finally, a couple of comments regarding the blog and wiki that were set up. Ideally they should have been arranged with some more anticipation. More importantly, however, is that they be made more general ‘artificial life community’ blogs/wikis and not just ECAL2007 ones (e.g. http://artificiallife.wordpress.com/ ?). That’s only a minor matter of changing domain name and title I guess. Perhaps it should be done for the wiki. On a different but related idea, would anybody (else besides me) find useful a user-submitted voted-on and commented-on paper repository in the form of an ‘artificial life community digg-style’ site? Just a thought.

ps. as the astute reader will have noticed (I always wanted to say that phrase) I am making an effort to not so easily refer to the brain or the mind as the place where things ‘happen’ when we ‘think/learn/remember’ things (e.g. the title of the post). I’m sure you will have found it utterly silly and downright annoying. I don’t blame you. But I think it should be relevant to have a situated, embodied and dynamical systems perspective not just in our research but throughout our lives, down to even our most meaningless day-to-day expressions. Although I understand they are not meant to be accurate, I reckon the accumulation of these phrases make it harder to turn cognition into a less GOFAI-oriented science. I’m sure, however, that for the larger proportion of the three people reading this blog it will not come as a surprise that their body is relevant to what they learn and remember.

Quick note before I begin: it just so happens that as I wrote this post the disk on my G5 died. I don’t think I have lost anything major, as most of it is backed up. Some of the most recent pictures that I have taken may have gone, because I had failed to backup those in some time (a couple of months maybe). I am re-backing up what I have to a third disk to avoid having the just one copy that I have at this instant, and will seriously consider buying another disk to have more backups. On to the original post!

§0 There is little point in making up further excuses - all I have time for these days is making this sporadic updates to the blog after increasingly large periods of absence. As in previous posts, I’d like to recall what I have been up to lately and what the plans are - mostly for my own reference in two months from now when I start worrying about not having done enough.

§1 So first of all, I’m glad to be very involved in this upcoming ECAL, which increases in its promise to be quite exciting (e.g. the format, the organisation, the keynote speakers, the one-track session, the workshops and associated events, and of course, Lisbon). Here are some of the things I have become involved in.

§1.1 I had the honour of participating as part of the programme committee in the paper-reviewing process. This is one of the best research-related learning experiences. I have mentioned this previously in this blog. Even though the quality of the papers in my lot was rather poor, the experience was still a fruitful one. Attempting to get to the motivations of the contributions and be constructive as to how to meet the established claims, is something that I must practice more often, as it is all too easy to simply throw the towel and discard from merely pointing out the flaws. The obvious benefit of this practice is that one becomes increasingly aware of the need to step into the reviewer’s shoes (or any reader’s shoes for that matter) when writing.

§1.2 Which brings me to my second point… I submitted a paper which was accepted for a talk. The work follows on directly from the things I have been working on for the last two years: it deals with the analysis of a situated agent evolved to perform an associative learning task requiring re-learning. This work compliments well similar work and most importantly (which is the reason that I put all of the effort in) fills an important gap in my thesis.

§1.3 Even more exciting still have been the collaborations on two other papers that were also accepted.

§1.3.1 One of the collaborations arose from conversations about the meaning of autonomy with Tom and Nathaniel following from Life & Mind seminars. Our contribution points out two different dimensions in which autonomy should be approached: constitutive and behavioural. Pointing out that most progress in the artificial life sciences has been achieved towards the behavioural side of it, and less so on the constitutive aspects. We also point out the serious limitation of methodologies that assume the unit of agency a priori (such as evolutionary robotics) in advancing any further along the constitutive dimension.

§1.3.2 The other collaboration arose from discussions with Peter about his work, which is closely related to my own. The contribution explores situated agents evolved to learn about their sensori-motor configuration. The idea of the work is to provide a existence proof as well as to analyse successfully evolved systems that weren’t given a priori parameter-changing mechanisms for a variation of the visual inversion task. The work differs from previous work and my own in that what is being learned is not a feature of the environment, but a feature of the agent’s body.

§1.4 The workshop Eldan and I are organising is going well. We received some submissions, but not as many as we were hoping for. There are some plans to possibly merge with Andy’s workshop on Neuromodulation, given the relatedness of the topics. I’ll update on this soon.

§1.5 Finally, as regards ECAL, I’m also really looking forward to the associated-event / workshop on “Dynamical Approaches to Development” being organised by Rachel Wood and others, for which I was honoured to have been asked to participate as a commentator.

§2 On a different note, I have been picking up on my reading. I have recently finished reading James Gleick’s “Chaos: making a new science” and Eric Kandel’s “In search for memory”. Also read Charles Dicken’s “Great Expectations” (which I had been meaning to do since I was about 14 - but I think my English hadn’t been up to it until quite some time ago). Currently reading Alva Noe’s “Action in perception”. Have also been reading bits and pieces from Maggie Boden’s wonderful and relatively new book (which I really want to get but is too expensive!): “Mind as machine”. There are at least 10 more on the ‘hot’ queue awaiting..

§3 On the more administrative side of things, I have also been involved in a couple of projects.

§3.1 I have been taking a slightly more central role in the organising the Life and Mind seminars this summer term given that Tom is away in Sweden. Although very exciting it is such a time-consuming task. Unfortunately, I am most likely not going to be as involved from now on, for all the obvious reasons.

§3.2 I’m also very excited that Ezequiel has given me the opportunity to suggest, put on offer, and now that it has been taken up, co-supervise a thesis project for the EASy masters course. This is exciting work on the analysis of the dynamics of slight (and hopefully interesting) variations (mainly to the transfer function) to small (one and two coupled nodes) continuous-time recurrent neural networks.

§3.3 More in general, I reckon having moved to the Informatics building since I arrived from Indiana has been very beneficial because of increased interactions with people whose line of work I am interested in. Being next door to the EASy master’s lab has meant that some EASy students come and chat with me about their projects, I find this extremely interesting. I also run into people like: Anil Seth, Adrian Thompson, Thomas Nowotny, Phil Husbands, Mike Beaton, Ezequiel and Inman much more often. This is great too.

§4 That covers mostly what I have been up to for some time. What I’m getting busy with now is:

§4.1 My third and last annual review, for which I hope this post will serve as an initial step in getting that done.

§4.2 Generating a full-fleshed grant proposal, following from an accepted pre-proposal. More on this later, hopefully.

§4.3 Marking NSAI programming projects. Not much to say here. I think I would be slightly more enthusiastic if I had not my thesis to write. More generally as regards the teaching of the seminars for NSAI this last summer term, what happened to attendance!?? and how worried should I get about it? I don’t feel like I did such an awful job - I brought as much excitement as I generally assign to the topics covered, which is a lot! I encouraged them to discuss as much as possible. The class was on the early side of things, at least for students in England: 9am (if they only knew I used to get taught mathematics and physics at 7.30am in college in Venezuela on two hours lectures). When I asked the students and some colleagues of mine, most would say that this was the reason.. but I’m not convinced. I would like to hear honest feedback from my students, I will ask around if a form can be sent around for it - to see what went wrong.

§5 The important thing ahead is thesis writing - which I am now more anxious than ever to dedicate my full attention to.

§6 One thing I hope to do more of, which does not reflect at all in what I have been doing, is increasing my interactions with the biologists. I had the chance to meet with some of them working on learning and memory in the pond snail Lymnaea stagnalis the other day regarding the grant proposal that we are putting together, and it dawned on me the difficulty of crossing the gap between the two fields. Although extremely daunting, I hope to overcome this gap in the following years of my career in a way that doesn’t entail me ending up as the “computer scientist that’s programming for the biologists” configuration - which is what seems to happen very often. I want the high levels of abstraction, the enactive, evolutionary, and dynamical systems approach to feed into work in learning and memory in Biology, without having to model every detail of two particular cells, for example.

§§ On a final note (and although this should be a research-mostly blog), I can’t help to mention that I have been extremely worried about my home country lately. The direction it has been taken towards during the last 9 years has been, in my opinion, an extremely inappropriate one. But what has happened during these last months has just been absolutely unacceptable. This has nothing to do with left, right, up or down political positions; and everything to do with one men seeking full control of all powers. Venezuela has fallen into the depths of a totalitarian regime and this is no secret: if you happen to like his line of thinking, great; if you don’t, then you must leave. I won’t go into it in any detail here because I fear the few readers that might be interested in the rest of my blog will simply not care. What I will mention is what is most frustrating for me. First, that I don’t know what I can do to make the situation better - I go absolutely blank when I try to think of things to do. The other thing is that I would be much happier dedicating my full attention to the things that interest me (my scientific research) and not the political sanity of my country (which is every day less mine!).

This post is all over the place: from the technicalities of hard disk failure, through paper-writing, paper-reviewing and teacher-assisting, all the way up to political nonsense. There it is.

ps. many many thanks to our mac administrator Christian for all his excellent support in the disk situation, as usual.

ps2. I didn’t manage to make space in my ECAL paper for the acknowledgments… very cheeky I know. So,

Acknowledgments. I would like to thank Peter Fine for proof reading and making suggestions to improve the English of the last version. If it were not for him (and Inman of course) the paper would read more like this blog! Also thanks to Hiroyuki Iizuka for words of encouragement about the value of the contribution when in doubt.

I have made some progress in the project that I have been talking about in the last posts, but it has also taken a lot more effort then I initially thought it would. Although I am enjoying very much doing new exciting research in to all of those things, the extra effort has made me realise that I should step back from this project and get back to the more practical aspects of writing my thesis, at least for a while until things disentangle in my head.

Before I move away from this I will make a couple of notes on the fronts that I am leaving open:

0. Evolve agents on two tasks: (a) pure thermotaxis on a simple two thermal-peak 1D environment and (b) thermotaxis as well as variety of behaviours in the same scenario. This has been quite easy, but I only realised recently how much precision I would have to add to the time-step of integration. Given that I am selecting for systems to be sensitive to initial conditions, I am more or less selecting for instabilities, and if they are more easily found in time-integration errors than in internal dynamics, then artificial evolution will not be kind to me. I usually use a time-step of integration of 0.1 - an order of magnitude smaller than the smallest time-constant allowed. I have had to make the time-step of integration yet an order of magnitude smaller: 0.01 - making evolution very very slow. I have also been trying some experiments where the first 1000 generations are evolved with a 0.1 time-step and then the last 100 generations with 0.01.

1. Reactivity measure and comparison between agents evolved for the two different tasks. I’ll update on this part later.

2. Analysis of the internal dynamics of a best agent that performs task (b). This has been the most interesting part so far. Because the system is non-autonomous, I have been looking at the set of different dynamics that the agent has when the temperature (environment) is fixed and how they change from colder to hotter. For the same agent I have seen how some phase-portraits go from simple periodic orbits, to something that looks awfully close to strange attractors, back to simple cycles. Most interestingly, the most chaotic-looking orbits happen near the more important decision-making regions: the valley between the two thermal peaks.

3. Measuring chaos. I have been reading and implementing several different measures for how chaotic a system is. The goal is to input a CTRNN with certain parameters and generate an index of how divergent and sensitive it is. Although I keep hearing from everybody how easy this should be, I have run into a number of problems. In fact, this has taken me the last 4 days entirely. I have focused on implementing an algorithm to calculate the maximum Lyapunov exponent. The problem here has been mainly to make the technique as automated as possible so that I can generate the measure as I vary one of the parameters (i.e. temperature). But each system can have multiple attractors, with orbits of different periods, and so on, and this generates some trouble for the more manually tunable parameters of the measures that I have been looking at.

I think it probably hasn’t been as bad as I am putting it now. I probably have this feeling just because I am tired from working on it intensively these last days - in the middle of having to move offices as well.

First some good news, the paper analysing the Hebb learning CTRNN without plastic weights got accepted for oral presentation to IEEE Alife in Hawaii. The reviews were not bad either. Inman thinks we should further that work and convert into a journal publication but unfortunately I don’t see that there is time to do so anytime soon. Surprisingly no complaints on the English - I have mostly Peter to thank for that. Thanks! I am actually very much looking forward to going to that conference and getting some feedback on that work.

The plan is that I will dedicate the months from January to September 2007 to writing my thesis. What this means is that all of the ‘new’ ideas that I will be coming up with I will have to write down as best as possible and put to one side for tackling later. Failure to do so will result in either forgetting them or even worst working on them and delaying the thesis-writing process.

There are two important things that I must finish while I write, however. First, the work that I did over the summer in Indiana. That work is close to being finished but I have lacked the time to see through the end of it. The second thing is the workshop for ECAL, if it gets accepted (fingers crossed).

While getting the usual ‘negative shaking of the head’ from the Mac user login as I type my password incorrectly I wondered, why do we not yet have operating systems that we can train by positive and negative reinforcements? A gentle pad on the side of the monitor (in resemblence to the sony aibo with its padding zones on the head that you can use to let it know that it has done something good), or if you are working alone at home then why not activate the voice recognition system and let it know that it has been a “good boy” after it has open the skype software in reaction to your wife writing you an email that she has arrived home and is ready to speak (say you are away doing research abroad).

Ok. So I’m guessing the reinforcement is the easy part. The interesting part is getting the computer to suggest new things such that they can lead to positive reinforcements more often than negative ones. But surely we could learn something from how infants, dogs, etc. do it. Who knows, we could even learn something about learning in real organisms as that gets done.