Difference between revisions of "CSSS 2008 Santa Fe-Projects & Working Groups"
From Santa Fe Institute Events Wiki
|Line 369:||Line 369:|
Let's plan for this group to meet on Wednesday morning (June 11) in the open timeslot at 9AM at the Peterson Student Center. Meet outside main lecture hall and we'll find an empty room.
Let's plan for this group to meet on Wednesday morning (June 11) in the open timeslot at 9AM at the Peterson Student Center. Meet outside main lecture hall and we'll find an empty room.
===(Bayesian?) learning rates in biological and social systems===
===(Bayesian?) learning rates in biological and social systems===
Revision as of 06:06, 12 June 2008
Antagonistic interspecific interactions
After chatting with several people about host-parasite systems and hearing some of the comments at the icebreaker, I want to see if others are interested in potential projects in this area. As a way of getting some brain storming started, I’ve just typed up some topics (these include ideas I’ve heard from others here at Santa Fe) to see if there is critical mass and an interesting question.
Topics (no particular order) Do we have one or two focal questions? Is there a desire to combine?
- Impact of host heterogeneity (= consumer-resource dynamics with identity issues): (Sarah writes:) Most ecological models of consumer-resource interactions assume that all consumers "view" resources the same way, i.e., each resource has only one possible phenotype. For a host-pathogen system, this means that all hosts agree on which strains are identical and which are different, since all hosts are targeting the same antigenic sites (epitopes) of the pathogen in their immune response. When pathogen strains compete in this environment, a broad range of cool dynamics result (Gupta et al., Science, 1998), depending on the strength of cross-immunity. There is evidence that hosts do not mount identical immune responses when challenged with the same strain of pathogen. In other words, a pathogen's phenotype is a function of the host. How does heterogeneity in hosts' immune responses--this multiplicity of phenotypes--affect competition among pathogens? These could be important results for the field. I'm thinking of doing some simple nonlinear dynamical analysis that builds on the framework in the Gupta paper. This problem seems broadly extensible to antagonistic interactions more generally, but I can't think of specific biological examples. Anyone interested? (Talk to me or post here!)
- Impact of immune system on host-parasite (/pathogen) interaction
- Seems like you could add some details of host genetics and then make up a matrix that describes the fitness dependences of the pathogens for each host genotype. -Devin
- Interesting paper by Recker et al (2008) that could be interesting to discuss on this track as well. -Devin
- Can the Recker article be modified by including extra host compartment to represent different host genotypes?
- I'm not sure we need to add extra compartments, especially if we use a status-based approach with the ODEs. It would be worth talking about this problem in front of a blackboard. How about Thursday evening or Sunday? -Sarah
Let's discuss 6pm Thursdaycompleted
- Let's get some pictures of what this would this system would look like. I'll try and add something, but other's should drop in a pdf/powerpoint/artististic rendering
- Pathogen Modularity I would be interested in modeling whether some aspect of pathogen modularity-based evolvability (e.g. "reassortability" or reduced evolutionary constraints between epitopes) significantly affects the evolutionary success of the pathogen. I'd love to incorporate realistic-ish parameters to get a sense for whether this kind of evolvability has a significant effect on evolutionary success, and if so, whether this significant effect is large enough that we might expect second-order selection for pathogen evolvability. --Molly
- First stab at getting some of our ideas in picture form: AII_Modularity
- This is very interesting. One could imagine using a (weighted?) network picture to capture the transitions. -- Jacob
- Right now we have a non-weighted network, but expanding the network to allow variation in edge weight will be a nice extension.--Molly
- Need homes, or just ideas for later
- Effects of pathogen competition on epidemic outbreaks
- Spatial heterogeneity of transmission of parasites-- If we have time, I'd love to put the model we have started to develop on a graph or something...see the effects of spatial heterogeneity and the structure of the population on pathogen population, and maybe see how dynamics differ as one moves from an area of homogenous viral population to a "contact zone" where two strains compete...-Molly
Infectious diseases: epidemic outbreaks vs. endemic steady states Direct vs. vector transmission of parasites/pathogens Non-genetic transmission of disease resistance
Interested? Please add ideas that you find interesting or would like to explore more (or just your name).
I'm not at all a biologist, but perhaps there is some way to quantify modularlity in terms of clusters in these networks? Look at cluster dynamics? Clustering that takes into account link weights (e.g. greater link weight than expected by chance)? Perhaps some of this has been looked at? As in http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=1865589 -- Laura
Epidemiology in general. -Ruben
Biological Levels / Phenotypes Discussion
We have a number of folks here either interested in or studying biology at various levels. I am interested in talking about ways in which it makes sense integrate different levels of biological knowledge into a representation of a system. For example, how might microRNA predictions be combined with gene expression networks (or proteomics or SNPs) to lead to a phenotype.
I am also interested in questions of how phenotypes are defined. Within an organ state (e.g., disease or not) for example, a phenotype might be defined as a gene expression pattern, a growth rate, a panel of microsatellite lengths, or functionally by in vivo or ex vivo capabilities to self-renew, etc. If what we are trying to understand is a larger question of disease or functionality, which phenotypes are interesting and useful (and possible!) to use?
I think these questions can be approached from a variety of ways. Off the top of my head, perhaps multi-scaled modeling or examining the system as a multi-level evolutionary system... I'm sure there are many others.
If you are interested, add your name and we can set up a time / place to talk about these and related questions. Also, please add anything you might want to include in a discussion!!
Something in Neuroscience
I (Nish) would really like to do some more intensive, deep research in Neuroscience (Computational being my perspective). While I am fascinated by neurological and behavioral diseases, I would be open to any kind of neuroscientific problem. Anyone else?
Well, I guess that's my cue! I did my Ph.D. in Neuroscience, and I'd be glad to impart what knowledge I have. What I'd really be interested in is seeing what the interesting problems are which intersect neuroscience and complexity studies. It's surprising, given the obvious complexity of the nervous system, but not many of the articles I read really use this box of tools.
If other people express interest, I could organize a tutorial or small working group where I talk about some of the issues I know about in neuroscience and how they may relate to complexity. (One off the top of my head involves storing and retrieving memories, and is done by Carlos Brody at Princeton; another possibility is applying Network Theory to functional connectivity using some new MRI-based data methods (including Diffusion Tensor Imaging)). But more generally, I think we could expand it to be a "speculative neuroscience" discussion, in which people throw crazy ideas at each other. Sayres 18:34, 4 June 2008 (MDT)
- Sounds like fun to me. I think Gerald Edelman and Walter Freeman have done some work in this area, actually, but let's chat (Chris)
- Yeah, I know (some) of Freeman's stuff, but not Edelman's; I just looked him up online. This would be a good opportunity to catch up on some of these ideas. I was introduced to the Freeman ideas earlier, but he hasn't been doing much original research lately, and a lot of the attention is more on his collaborators, like Gyorgy Buszaki (sp?), whose work I would consider more focused on the trees than the forest. :) I myself have studied things at a slightly lower level than Freeman's big-picture ideas, but I think that's what this course is for. I will bring also some work I think is relevant in using forms of network theory to find clustering of objects (as well as in the olfactory system of insects, e.g. Gilles Laurent), and maybe talk about some newer models of decision making.
I see Nish has a little note below; I'll put some of my responses below there. I'll also go ahead and propose a small tutorial for next week; my plan would be to spend a little time presenting what I know, and encourage people to discuss their own interests. Sayres 14:38, 6 June 2008 (MDT)
- yeah, wish I had my books from home here, my memory of Edelman's off the top of my head, but he's real good at synthesizing things (his 2001 book 'a universe of consciousness' is pretty damn good at that, once you get past the popsci intro chapters). Also worth checking out is Peter Gardenfors, and I really like Eric Baum's 'What is Thought?', which though based in computation rather than neurosci, is also great at bringing together so much new research. Terrence Deacon and Merlin Donald's stuff is also excellent, at least on how the evolution of the brain can shed light on these issues. But yeah, let's set up a time to chat on all this.
- This discussion about the storage of short term memory is very interesting. I am interested in finding more realistic models of decision making, e.g. how to model an exploratory process that may eventually lead to a decision? I'm mostly thinking at how to apply these ideas when doing agent based models of social systems. Broad ideas on the architecture of memory in the brain are definitively invaluable! Also, I'm very good at throwing crazy ideas on! Count me in! Giovanni
- Hey Giovanni, sounds good! If you're interested in decision-making, I'd like to direct you to the work of Greg Corrado, a friend and colleague of mine at Stanford. He's done work (with Leo Sugrue and Bill Newsome) on decision-making in tasks where the odds of a choice leading to a reward update dynamically. Here are links to a few of the papers:
- short review article outlining general approaches towards modeling decision-making in neuroscience: a good place to start
- Science paper presenting the initial experiments on a simplified foraging task, and trying to understand neural correlates of perceived "value"
- theory paper explaining the "LNP" (linear-nonlinear-Poisson) model used to model decision-making processes
I'll try to talk about this in the tuorial on Wednesday. But in case I don't have time to do it justice (likely), I think their model provides a nice compact description that does a very good job of explaining some complex data. Cheers, Sayres 17:35, 9 June 2008 (MDT)
UPDATE 06/11/08: The tutorial went well. We had a lot of good conversations and questions, and I think we may have even learned something. We decided to continue the discussion, in the format of a working group.
There seem to be two groups of people at the tutorial: those who are interested in working together to develop a project related to one of the questions we discussed, and those who want to know more background, but not necessarily a project. I propose we have a second meeting which is divided into two sections, one for each group.
I will update this space with a suggested time shortly. Sayres 13:38, 11 June 2008 (MDT)
I'd like to do anything relating to the introduction of misinformation into a system, but one concrete suggestion is looking at how one might evolve a skeptical response to defend against being "defrauded" (this could be in a social science or biological system).
if you can't grow Collapse, you haven't explained it
Jared Diamond describes a five point framework for collapse of societies. These points are:
-resilience of the environment to human caused damage
-society's response to its problems
My proposal is to test these points in an agent based setup. we could for example use the parameter sweep etc.
If you are interested, add your name and we can set up a time / place to talk about these and related questions.
Although fun to do I decided to go for a topic from which I could learn more, feel free to go ahead with it if someone is interested. Dirk
I am also very interested in this Richard
Could we adapt this and include an urban perspective? Flávia
Everybody needs his physicist. :-) - Ruben
I'm interested in this project as well. Steve
I am interested! John
Heck yeah! Rory_Sayres Sayres 18:35, 4 June 2008 (MDT)
sounds fun. hope I can be of any help... Francois
Let's make networks secure
I'm interested in making 'secure' networks with bound resources (like it is normal in reality). This is a very general questions and has many oppurtunities. For example:
- I have an epidemic and not enough vaccine - who should be vaccinated, who not?
- Terrorist try to smuggle a bomb into the land of Oz. Which flights between which airports should be especially watched to minimize this risk?
- How can we try to secure the internet with special anti-virus hubs? Is it possible to stop the epidemic of computer-viruses by special 'antibody' servers?
This just came to my mind and if anybody has other good ideas or wants to comment this - please feel free.
- How to select nodes in the water system to detect the pollution efficiently? - How to select individulas in the social network to let advertisement to spread efficiently? - How to select blog to let people just see a small part of blog and get more information? Data mining will be useful to analyize huge dataset. [see http://www.cs.cmu.edu/~jure/pubs/detect-kdd07.pdf] -Jiang
- Might be interested to see if all these problem can be generalized into a single combinatorial or asymptotic one. What kind of tools would you like to use? (I guess not agents :P) Giovanni
Network Security Concepts Inspired by Biological and Social Systems
Ruben – Was about to post this when I saw your entry. Sounds like we might have some overlap in our ideas…
One area that I am interested in for a project is applying concepts from biological and social systems to the area of computer network security. This is a broad topic and I am hoping to generate some discussion that might lead to a more well-defined research area.
In the biological area, for example, there have been recent developments in Artificial Immune Systems which borrow techniques from the immune system that enable virus detection and elimination in a self-organized and distributed manner. In particular, Stephanie Forest here at SFI has done work in this area. (see http://www.cs.unm.edu/~forrest/ , http://arxiv.org/ftp/arxiv/papers/0804/0804.1266.pdf) Other interesting research has been done by John Doyle at Caltech in which he compared the “robust yet fragile” organization structures of computer and biological networks (see pages 96-111 http://www.cds.caltech.edu/~doyle/GENSIPS/GENSIPS.pdf , http://www.cds.caltech.edu/~doyle/CmplxNets/)
In the social science area, I have been thinking about trust relationships and how they could apply to computer network security. Trust relationships are well established in computer networks for public key certificates. However, to my knowledge, there is no way to look at the “trust” of pieces of data in a network. Data is often scanned as it enters a network but is not tracked once it is inside to ensure that it behaves properly. This leaves many networks “hard on the outside, but soft and gooey on the inside.” One idea might be to leverage trust/reference concepts in social networks (e.g., eBay, citation networks, Amazon referrals, social websites, etc.) to construct a framework for “trusting” data throughout its lifetime in a network. For example, the more frequently that a piece of data is used effectively by an application might increase its trust. Also see http://www.mindswap.org/papers/Trust.pdf
Please list your name if you have any interest in this topic. Thanks.
I'm happy to help look at this topic, but I suspect that for proper interdisciplinary work you need experts in biology, not computer systems... -- Laura
Incorporating Data into Agent-Based Models
The rise of the "omics" fields of biology (e.g., DNA data in genomics, protein data in proteomics, etc) have resulted in a bewildering mass of data. I'm interested in exploring how these data can be incorporated into agent-based modeling strategies. Perhaps the data-mining folks have similar issues?
Perhaps this is a naive question and someone already knows about an instance where this has been successfully accomplished. If so, please forward it along to me!
In any case, I'd like to do anything from just talking about this problem to creating a simple agent-based model that uses publicly-available data. For example, a simple model of bacterial or yeast growth could be coupled with gene expression data from the cell cycle. I'd be very happy to explore any other systems folks are interested in as well.
- I understand nothing about biology, but I'm always dealing with empirical data and the challenge of incorporating it to agent-based models. So, maybe we can talk a bit about it. (Flavia)
A Dance Evolution
I'm interested in trying to take Liz Bradly's alphabet dancer (that we saw Monday evening) and seeing if anything interesting/aesthetic could be done in a context where a set of such modeled dancers evolve their choice of movement (and perhaps their location/orientation) based on what their neighbors are doing.
My first thought was to evolve the selection of agent dance movements by analyzing how each selected movement (or perhaps simply their hand positions) temporally/spatially relate to the selected movements of its neighbors.
It might be an interesting context to explore self-organization and the role of conflict and cooperation in producing interesting emergent properties.
I'm thinking the primary work would be done in Netlogo. But I've noticed that Maya can be downloaded for a free 30 day trial period so the result could hopefully be visualized in 3D.
If you're interested leave your name below: Steve
One thing you might want to look at is looking at how dance notation (i.e. formal languages for describing choreograpy) could be used as a basis for "mutation", analogously to mutations on CAGT... Laura
(Giovanni): With the term social media one can indicate the various web 2.0ish communities that are popping out on the internet these days. The following ideas come from watching Wikipedia's users community but I think there are correspondents in the other major social websites, as well on non-internet based communities (e.g. networks of scientific pubblications).
- Ownership of encyclopedic articles. Wikipedia's policy is that no one can claim ownership on a wiki entry. However, some forms of ownership are sometimes tolerated e.g. when an expert on a certain topic imposes his autority on non-expert editors. While it is generally wise to let the experts user to do this (since you want experts to collaborate to the project), sometimes this can lead to pathological cases in which the "owner" dictatorial methods discourage any other user to do any edit at all. This is a sorta of prisoner's dilemma, since you do not want the expert users to be banished by the community just, but at the same time you want to keep it as open as possible.
- Vandalism. Given a model in which agents either change the content of a page based on their point of view (these concepts have a precise definition) or revert it to a previous "clean" version, how can a community fight off vandal users? I would like to explore this problem with a "neutral"-like model e.g. in which the social phenomenon of vandalism is just explained in terms of "distance" from a point of view (again, these concepts do need and have a precise definition) and a "culture dependent" model, in which vandals form a population on their own and thus a "vandalic" cultural traits exist and is clearly identifiable by the agents.
- Community dynamics in terms of double selective pressure. This is a generalization of the previous two. These kind of problems suggest that there is a double selective mechanism by which users are "selected" into the community based on its current status (whatever this thing is), and at the same time the community's status is influenced by the users that live in it. A quick example: a Wikipedia plagued by too many vandals would probably discourage the average user to get in and fight vandalism, which makes life easier for vandals etc.
I would like to explore these problems either with multi-agent simulations or with differential equations. Data are usually not a problem if one wants to study social websites with open API.
On the other side of the MAS-coin, I'm also interested in general methodological problems of Multi-agent systems modeling.
- "Evolving" interactions in MAS. I talked briefly about that during the brainstorming session of Wednesday. I think some people were interested in that, so better to talk directly.
- Causality in MAS simulations and network motifs. Here the idea is to look at/develop algorithms to extract network motifs from multi-agent simulations. One has to use a model in which it is already clear what a causal relationship between two events is (I call these also interactions), or at least use the concept of "cause" used in graphical models. Other big question: once one has this kind information, how to use it?
I would be interested in pursuing this. There are some good ins'ights on how this was 'managed' in the development of Linux in Raymond's Cathedral and the Bazaar. http://www.catb.org/~esr/writings/cathedral-bazaar/cathedral-bazaar/ Talks about that need for the 'inner circle'
what kind of questions do you want to solve using multi-agent model? Actually, I am really interested in the web2.0 and information diffusion pattern in this kind of web? When is the next meeting? Please add me in this discussion. Jiangwu.firstname.lastname@example.org...Jiang
@Craig: there is also an interesting video of the GoogleTech talks series titled "How Open Source projects survive poisonous people". The conclusions the speaker draw out of the talk go pretty much in the same direction, which makes me suspect that there is some empirical phenomenon worth to study.
@Jiang: I am interested in quantifying the level of health of a community tha collaborates over a social media, and a good proxy may be the "quality" of the content that community puts in the media itself. A multi-agent simulations may give a decent evidence of interesting phenomena going on, but I am also interesting in other modeling tools of course!
What about discussing this stuff over a cup of tea tomorrow at SFI? Giovanni
Financial crisis in Housing Markets
UPDATE: Follow the link to this project: Housing market crisis
After hearing the questions from Dan about market crashes, I got curious about a more recent market crisis, the housing market/home mortgage crunch. Over breakfast a few of us were discussing this and came up with a few questions.-Devin
- Is this really a complex system, or just something that is not transparent?
- What really caused our “crisis”?
- What happens if the “Fed” had let Bear Stearns go bankrupt?
- The crisis seems to be rooted in a separation between good information at the people who benefit from that information. (Abby)
- Can this be modeled in terms of banks with different risk behaviors/practices? (Devin)
- What about creating a simple ABM model with borrowers and lenders on Netlogo so we can play with heterogenous strategies from the perspective of the firm?(Carlos)
- We can introduce multiple lenders and introduce competition (or collusion?). (John)
- There should be accessible data on this. (John)
- Let's refine the idea. (John) see next section
Back of napkin model
|Borrow Agents||Lender Agents||Output|
- Hey, just some thoughts on this. Mark Taylor's work on micro/macro finance and crashes in light of complexity economics might be of interest here. He does a great job of explaining the LTCM/Asian tigers crash of 1999. David Harvey's work on the redistribution of crashes via international financial networks as ways of managing accumulation crises also has a lot to offer complexity economics. Eric Beinhocker's work also does a great job of laying out parameters of complexity economics as a whole. My concern about modeling these types of crashes is that modeling a given system can be really decieving, cause really there's an intricate web of complex nets at multiple levels of scale and widely spatially and contextually distributed, all in careful equilibrium, such that minor disturbances in one can trigger a phase change in another, making endogenous analysis problematic. Even the time series analysis done by Mandelbrot and co. in stock markets really does little more than a form of what Liz was doing with expanding 1D time-series of a 3D system, but with a LOT more degrees of freedom at stake. Which isn't to say this sort of modeling doesn't interest me, I just wonder how to do it justice. So, with the Bear Stearns crisis, it seems to me we've got another example of the financial industry coming up with new 'financial products' (in this case, sketchy loans) designed to avoid the fitness conditions provided via market regulation in congress. This leads, as it does everytime the financial sector outwits the regulators (as they do cyclically), to overleveraging of credit, inflation of the market, confidence crisis, government bailout of those who created the bubble, and large scale wealth transfer from low end of the market to the high end. Don't get me wrong, some lose their shirts on the high end, but overall, this sort of 'managed credit crisis' acts as an overall wealth pump from the low end to high end of the market, not due to conscious planning of the crisis on the part of the financiers, but as a macro effect of micro-rules based on outwitting regulators to maximize short range profits. Problem is, though, so many other networks are nested and intertwined here, I mean, how do you take into account shifts in the fitness environment leading to the evolution of new financial technologies, all of which caused the setup for the model in the first place? This problem REALLY interests me, just not sure how to do it in a way that doesn't ignore these complexities. Ideas? Chris
Notes from the margin (6/8/08 - John)
We need to keep this fairly simple. We're not trying to model the US economy. Let's stay focused on what questions we're trying to answer.
Two types of agents: borrowers and lenders.
A borrower agent is described by its credit score. A credit score is an weighted aggregate of the following four attributes of the borrower:
35% — punctuality of payment in the past (only includes payments later than 30 days past due) 30% — the amount of debt, expressed as the ratio of current revolving debt (credit card balances, etc.) to total available revolving credit (credit limits) 15% — length of credit history 10% — types of credit used (installment, revolving, consumer finance) 10% — recent search for credit and/or amount of credit obtained recently
Credit score distribution in the US is:
As a first-order approximation, this can give us the distribution of “risk” profile among borrower agents. Additionally, depending on income (and perhaps credit score itself), value of the property being considered is probably a variable – that is, amount of the loan (certainly, the maximum one can borrow is related to one’s credit score). For the purpose of our model, we should just assume adequate supply of homes fitting the borrower agent’s borrowing capacity. And given the interest rate (and for some people, monthly payment), the borrower agent will either go for the loan or not (and shop around).
We need to also determine an agent’s propensity to borrow – what motivates an agent to decide to purchase a home, and how do we model that?
When a lender agent receives an application for a mortgage, it will process the application according to some pre-specified underwriting rules.
Some good information on underwriting practices can be found at:
Information on subprime lending and subprime crisis can be found at:
We need to come up with a simple algorithm for the lender agent in the model, which will determine if the loan will be made, and if so, return principal amount of the loan and term structure of the loan (fixed or variable interest rate and length of loan – and if the rate is to vary, rules for that). What are the key variables we can tweak to either encourage or discourage non-prime lending?
And after the loan, so what?
Certain percent of the loans will default. How do we model this? For the first model, we can assume some percentage of default rate, given the borrower agent’s credit score (hopefully, this data is available). When do they default (first year? Second year?...)? Can we make some simplifying assumptions (e.g., if you don’t default during the first five years, you won’t default, or something like that) for the first model – we can refine the model after we get the first, simple model working.
Future Refinements we can consider for the model
Lender agent’s ability to continue lending should be dependent on previous periods default history (some can/should go bankrupt), which should be directly related to its profitability.
Borrower agents will shop around (quality index for lender agent, based on the quality of its portfolio, or perceived "easiness" of getting the loan).
Dynamics systems applied in networks
A year or so ago I discovered this researcher who uses "Biologically-inspired attractor-selection" methods to route data through a network. It's been a while since I read it, but I believe the idea is to use systems of differential equations to solve networking problems such as finding robust, energy-efficient or shortest paths. I think it would be an interesting project to use both network and differential equations to derive some results, analytical, numerical and simulation. The network need not be a computer network.
Pessimistically, the approach might be just "bullsh*t and hype", but if so, we could analyse and suitably criticise it, and perhaps develop a better approach. Regardless, if anybody is interested, feel free to speak to me (my crap laptop broke recently so I won't be on the net as frequently as usual).
I think there are a few of us (notably Ruben and myself) that are interested broadly in bio-inspired network stuff (e.g., routing, virus spread, artificial immune systems, etc.) Perhaps we could try to consolidate our interests into a single project. Thoughts? - Justin
Perhaps work by Lidia Yamamoto would be of interest: http://cn.cs.unibas.ch/people/ly/ -- Laura
I think Kenji Leibnitz' work in bio-routing is conceptually very intersesting, but not entirely convincing from a practical perspective. But I would be happy to meet with people to disucss and look for a project in in this area. Laura
My computer works again, for the time being. I'm not too focus on biologically-inspired methods necessarily, though they are interesting. I thought this method would be interesting, but now I realise they're stochastic differential equations which don't lend themselves to analysis as easily, I believe. However, it may be possible to ignore the stochastic term. I will sleep on it. Paul
Evolving organizational flexibility in dynamic environments
Organizational adaptability and robustness in dynamic environments
Inspired by Josh's model of evolving organizational hierarchy on Thursday, over lunch a few of us discussed how this could be enhanced in several ways. I think that the 'business' in the model can be considered analogous to human societies which have used different organizational structures to cope with differing dynamic environments, and perhaps have corresponding weakness or strengths in said environments. Many geographers/archaeologists consider the level of organizational complexity and inter-linkedness a key part of understanding the resilience and flexibility of a given society.
Some of the ways in which the model may be enhanced are:
- Encode more sophisticated adaptive strategies
- Add a cost of changing institutional structure, perhaps some sort of sunk-cost effect?
- Add more realistic dynamic environments
It is clear from historic human/environment interactions that some societies found it very hard to adapt to changing environmental conditions, and one possible explanation of this is that changing the structure of a society is non-linear process with thresholds, and for vulnerable societies crossing this threshold level was enough to cause collapse.
We could use historic case studies of failure (Norse Greenland, Easter Island) and success (Norse Iceland, many of the other pacific Islands, Japan) and ask, given we know the dynamics of the environment (though palo-environmental reconstruction) if the structure of the society had an impact on the success (or failure, i.e. mal-adaption) of that particular example.
I'm interested: Steve
Me too. I'd like to hear about the cases: Cathy
I can see some parallels between this and the classic tracking problem in control theory. I'd be interested. srideep
Antony and I have discussed a related problem. There are some recent results in biology relating the fitness cost of developing a sensor system to the entropy of a randomly varying environment. The dynamic environment in Josh's model could be generalized to such an entropy-parameterized environment, and one might expect a transition to the hierarchical structure (which is, more or less, a sensory system) as the entropy varies. Jacob
Let's plan for this group to meet on Wednesday morning (June 11) in the open timeslot at 9AM at the Peterson Student Center. Meet outside main lecture hall and we'll find an empty room.
We want to use simulation model to study the organizational agility or robustness. I have find a electronic book. This book is about agile organzation, how to manage information network in the organzation from the viewpoint of complex adaptive system.(THE AGILE ORGANIZATION) -Jiang
Inspired by David Krakauer's lecture showing that evolution can be seen as a form of Bayesian learning, a question that naturally arises is how to measure learning rates in natural and artificial systems. In economics, learning curves and experience curves are well-documented, typically following a power law shape. Can we quantify progress in biological, technological, and cultural evolution/development using this framework? Can we use efficiency as a universal measure? Also: what is the connection with the MEST-compression idea (Matter, Energy, Space, Time) of futurist John Smart? (For example, think of future atomic neural nets mentioned by Alfred Hübler as a more compressed form of future intelligence than present state-of-the-art computing machines or brains). Let me know if you'd like to brainstorm some of these ideas. Béla
An agent-based simulation of a pollination network
Flowers need animals to visit them for reproduction, while these animals visit the flowers for food resources such as nectar. The pattern of interactions between flowers and insects can be thought of as a network of interactions. Different questions emerge in such network:
. What strategies flowers use in order to attract insects? What strategies insects use in order to visit the different flowers?
. What is the explanation for the specialization or the generalization of flowers and insects? why some flowers try to attrack a wide range of pollinators whereas other flowers specialize on a single species?
We would like to make an agent-based simulation based on an evolutionary game theory approach to try to find explanations for such questions and compare it with real data.
The pollination network wiki page
At first sight, it seems obvious what a structure is. However, there are a lot of different definitions of structure in use - in and between different communities.
In the first meeting, I (Tanja) informed the others about my motivations for looking at molecular structures based on RNA features: showing existing definitions of RNA structure -- for example, minimum free energy, consensus or abstract shapes, as well as discussing the importance of structure from a phylogenetic viewpoint.
Moreover, we have had a short discussion about the Allegory of the Cave by Plato and started discussing the following papers:
. Structure, Moshe Koppel, 1988
. Form and Content in Thinking Turing Machines, Oswald Wiener, 1988
Both in: The Universal Turing Machine / A Half Century Survey, edited R. Herken, Springer-Verlag, New York.
We ended up with a lot of questions e.g. about TM, what is a sequence, minimal descriptions, underlying structures, fundamental structures, representatives, or the use of definitions in general - e.g from a historical viewpoint, as well as to categorize them -- e.g. looking at the purpose.
We already have different viewpoints in our group (e.g. architecture, information theory, linguistics, logic, philosophy, computer science, biology). At the last meeting, we decided to start a wiki page to play with the different definitions between the fields and to prepare our next meeting.
The structure definitions wiki page
We have already made an appointment for the second meeting:
9.00 on Wednesday, at Suite 3 (number 39 on the map), upper floor.
Somebody else, e.g from physics?
I will bring what I can from physics.-- Craig
Geometry of Fitness Landscapes
As Dan mentioned today (June 9), the geometry of fitness landscapes is relatively unexplored terrain, with most of the underlying trait spaces being either simple (graph) or highly structured (Euclidean) . I've been interested in this topic for a while, so inspired by Dan's comment I thought I'd see if anyone else shared this interest. As a point of departure: there are some interesting articles in the Santa Fe volume on Evolutionary Dynamics, a paper by Stadler, Stadler, Wagner and Fontana (this latter regarding the underlying trait space, see also), and some recent fancy work using computational algebraic geometry (which I don't understand entirely: http://www.biomedcentral.com/1471-2148/7/60/). As a related question, one might imagine exploring the "measurement theory" of fitness landscapes -- how do we reliably reconstruct the landscape geometry from finite measurements?
Anyone interested? (Jacob)
Update! Since interest seems to be accumulating, let's have a first meeting tonight (June 10) at 8pm outside the main meeting hall (we can find a home depending on availability) to discuss the many ideas for an angle on this question. If this time really doesn't work, dash me an email or complain here! --Jacob
I am interested in the above, and I am familiar with how these questions have been addressed at the level of the individual protein. I'd be interested in learning more about models that apply to extremely rare monstrous phenotypic changes that create new species and things. If one has to choose a specific level, I feel that these two extreme ends of biological hierarchy are the most interesting places to ask this kind of question. But if there are fancy models out there that find some way to approach this question without specifying the specific level of hierarchy, I’d be interested in learning more about those too. -Molly
- I'd be willing to explore this a bit. I have some interest in algebraic topology, computational geometry stuff. However, it is not clear from the paper abstract what kind of 'computational algebraic geometry' they are talking about. srideep
- I might suggest looking at some of the work by Sergey Gavrilets as a stepping off point for some modern thought from biology on fitness landscapes. Poking around on the web to find a quick summary of some of his thoughts on Holey Adaptive landscapes I found this Lecture he gave. It's long, but the end slides go into the meat and the begin provides a nice evolution of the ideas. Devin
- On Thursday at Jeremie's viral modeling discussion, I'll talk about some work my collaborators and I did on modeling evolution in flu. It's a direct implementation of the theory of epochal evolution. We used neutral networks (inspired by stuff by Fontana, van Nimwegen, Stadler) to create a fake genotype-phenotype mapping, which allowed for slight or major punctuated phenotypic changes. I'll mention some of the problems with the GP map (we tuned Kauffman's NK landscape to make it, and converted the fitness levels into arbitrary phenotypes--unlike the standard models, we allowed to fitness emerge endogenously from the disease dynamics). It's just neutral networks, punctuated phenotypic change, and frequency dependent selection. I'll be clearer on Thursday. Sarah
- Also, some people who've done nice empirical work on the role of epistasis in fitness landscapes are Joe Thornton, Dan Weinreich, and Richard Lenski & co. I'm not up on all of Lenski's work but would be happy to discuss it. Can't unfortunately make the meeting tonight. Sarah
- I looked at the Lenski & co. stuff, at least the new (and fancy) stuff, a few months ago and didn't make much progress in understanding it. However, if folks want to try to read the paper linked above (and here), I'd like to understand it at some point. Jacob
On a related (but slightly less abstract) note, I am interested in modeling how interacting species co-construct their own and each other's fitness landscapes. I'm imagining an agent based model where each agent has "genes" which specify its location in a fitness landscape, and also other genes which add some sort of random function to the landscape (as a very crude approximation, one could imagine using a mixture of Gaussians). As the species evolve, you can watch how they move around in fitness space, and how they deform the landscape. This is an extremely rough idea, but I'd love to sit down and talk with some people about it. Maybe in a related discussion to one posted above.
Any takers? David
- So in populatoin genetics, I would call this coevolution between two or more players of an interaction (such as host-parasite, or plant and pollinator). Coevolution would occur "where the fitness of both species is affected by the distribution of traits in the other species." (Gomulkiewicz et al 2007). I might sugggest looking into some of the coevolution theory literature from such authors as John Thompson, Richard Gomulkiewicz, Scott Nuismer, Sylvain Gandon and Curtis Lively. These are folks who do model the reciprical effects on fitness of both players of an interaction. For instance, the selection on plant depends on the genetics of the pollinator and the selection on the pollinator depends on the geneticss of the plant. One player cannot remain static. Otherwise of course, there would be no CO in coevolution. Devin
- Stuart Kauffman developed a theoretical coevolutionary fitness landscape (NK model again), which he describes in the linked paper and maybe in his book, At Home in the Universe. I think the Fireside Lounge has a copy. Also, a Devin points out, the landscape in host-parasite models tends to be highly squishy (though rarely modeled explicitly) from coevolution of the pathogen with host immunity. The latter is obviously not strict evolution, though. Sarah
Something that came to my mind during the discussion of fitness landscapes was the idea of rate of adaptation to a dynamic fitness landscape. E.g. if the fitness landscape changes (e.g. climate) such that my current local maxima is now becoming a minima, how quickly must I explore the landscape to esape extinction (dinosaurs seem to have failed this :-) Possibly interesting for both biological (antibiotic resistance, invasive species; climate change) and social systems (evolution of social systems in response to major technological change). Perhaps not enough data to do sensible work.
Any interest? Laura
- I would say that invasive species provide a rich source of data on this subject as they can end up in an enviroment (and possibly a fitness landscape) that is quite different from their native range. A couple of competing ideas here is that invasives can have rather flat fitness functions with respect to their environment (lots of phenotypic plasticity) or perhaps have enought genetic variation such that an evolved response occurs in the new invaded range. There are of couse invasives that end up modifying the invaded range as well. Devin
- That's interesting. I was thinking also about the response of the invadee (if that's a word). The (local) fitness maxima it occupied is no longer a maxima, so can it adapt (re-attain a sufficiently high maxima) quickly enough to avoid extinction. Pehaps antibiotic resistance is a good example?
Looks like there's plenty of interest, or at least enough to call a meeting. When works for folks? Maybe sometime tomorrow? Jacob
Summary of Modeling Tools
A few of us were talking about how helpful it would be to have a summary of the common and useful modeling tools annotated with their inherent assumptions, best practices, pitfalls, etc. At the moment, we thought we would create a wiki page with this information. Please stop by and comment on your favorite tool. In the future, we may bring all interested folks together to discuss and try to come to a consensus.
Evolutionary Game Theory
A few of us were thinking that there wouldn't be enough time to discuss enough topics in EGT in a single tutorial, so I decided to post an offer for a working group that could meet fairly regularly to read and discuss papers from the field, suggest new topics, and possible projects. So far we've thought about looking at evolutionary branching models (in, say, a colony of yeast that produces an enzyme that can be shared by all individuals of the colony) and extending them from one population to two.
Please let me know if you'd be interested in joining. Feel free to add your name and topics you'd like to discuss.
Update! This has been scheduled on Friday from 3 - 5, location TBD.
- I'm interested! Actually, I know nothing about the topic, but I think that it could help me in my research. There is an interesting article about segregation and game theory (Zhang2004). Maybe this could serve as inspiration for a project. (Flavia)
- I'm interested too. Kathleen
I (Jeremie) propose to set up a workshop on viral modelling (biomathematics, epidemiology, virus kinetics, evolution, networks...). The idea would be that anyone could prepare a short and general introduction to its research area. Please let me know if you'd be interested in joining. Feel free to add your name and topics you'd like to discuss.
This sound like fun. All of the above are of interest to me. Alex
I'm interested - Paul
Great idea. I might have some info on computer network virus propagation that I could contribute. - Justin
Ok, I suggest meeting on thursday from 4.30. Let's say we could prepare a 15-20 min tutorial, just to see if we are on the "same page". How does it sound ?
Please put me on any lists for such meetings or tutorials. I am very interested! I have studied quasispecies theory some, but I have no hands-on research experience in any of this. Molly
Social (and other?) Networks
There are a number of people who are doing work with or related to social netwworks. Is there interest in a social networks working group? Alternatively, in a more general working group on networks and network based methods (biological, social, etc)?
Show interest and maybe times for a meeting below:
Laura -- I'm free this evening; maybe tomorrow evening before tutorial Jeremie -- I'm interested also; free tomorrow evening !
I'd also be interested in this. Tomorrow evening sounds good. Mark
Tentatively scheduled for 7pm -- Laura
Can we move at 6? I would like to attend Béla's tutorial ... Giovanni
Update (6/08/08): Ok, here's the link for Theory Group Texts page. Post any texts and links you find here so we've got a central clearing house. Some possibilities mentioned were Bruno Latour, Indian philosophy and complexity, etc. Also, next meeting, Tuesday, lunchtime, one of dining hall side tables. So far on agenda - trying to get some of the major terms from the Delanda (strata/network, double articulation, Bwo, agent/operator, intercalary agent, etc.) and discuss further implications of diagram/abstract machine and hierarchy/meshwork distinction.
The general will has spoken and set a rough agenda for our joint exploration of the weird intersections between continental philosophy, critical theory, and complexity. We'll start by reading selections from Manuel De Landa's A Thousand Years of Nonlinear History -- Introduction, Sandstone and Granite, Species and Ecosystems, Arguments and Operators, Conclusions and Speculations. There are several copies floating around (if you have one could you please post here and if you need one do the same). We'll meet Sunday afternoon, to avoid conflict with the hike on Saturday, specific time to be determined.
Update! First meeting 4:20 - 6:00 on Sunday, at the coffee shop/lounge area.
We've discussed several next steps, which I'll post after dinner :) --Hypothetical Hypochondriac
- Possible topics we threw around for further discussion - basic working intro to and/or discussion of any of the following thinkers that have been linked to complexity studies - Gilles Deleuze (rhizomatics, diagram/complex assemblages, virtual multiplicities, non-binary theories of language, desiring machines, nomadism vs. state structures, capitalism and desiring-production, major/minor sciences/social groups, networked models of mind, time and duration, intuition as method), Jacques Lacan (matheme-ization of the freudian unconscious, subject/language, master signifiers and social discourse production, ties to topology), Alain Badiou (ethics of the event, ties to cantorian set theory, mathematical ontology), Michel Foucault(decentered subject, disiplinary institutions, biopower, shifting epistemes), C.S. Pierce (process semiotics), Whitehead (process metaphysics), etc. Any ideas/suggestions? (Chris)
Spoken like a true volunteer, Chris! I would very much like an introduction to Deleuze. Peirce and Whitehead would also be fantastic. I would further propose that with many of these thinkers (Deleuze, Lacan, Badiou) we can have some very interesting discussions of the use, misuse, and abuse of technical/mathematical/complexity metaphors outside of the home field (or -- is there even such a thing as abusing a metaphor?) . How can such translation best be achieved? Lacanian mathematics would be an especially interesting angle in my opinion.
Another possibility: Leibniz and the "pre-history" of complexity science! Maybe reading the Monadology, etc.
I believe Tanja will post a link to Krakauer's Metahistory article, to support a broad conversation about History, laws of History, the dangers of such laws, etc.
Hey All- Ok, found some public domain texts, will keep posting them to my profile page, so far including some Deleuze, Badiou, etc. Chris
Thanks, Chris! I am looking forward for our discussions about Deleuze and De Landa and I would very much like an introduction to Lacanian mathematics. Futhermore, I would also be interested in discussing the Monadology of Leibniz. As mentioned in the last meeting, it would be nice to take a look at the following paper: The Quest for Patterns in Metahistory, by David Krakauer, SFI Bulletin. (2007). In addition, a link to the frescoes in the Cappella Scrovegni by Giotto, which I mentioned in the previous meeting.
One more link: http://www.netbase.org/delanda/meshwork.htm MESHWORKS, HIERARCHIES AND INTERFACES
Another related idea on the emergence of knowledge: "Empiricism and a Philosophy of the Mind" by Wilfred Sellars http://www.ditext.com/sellars/epm.html It might be helpful to skim the intro (argument against the myth of the given and related empirical theories) and then check out Sellar's myth of the "Rylean Ancestors"
Non-equilibrium thermodynamics and the production of entropy
I brought with me the textbook "Non-equilibrium thermodynamics and the production of entropy" by Kleidon and Lorenz and I have been trying to understand over the last 6 months what it is all about. The claim is that Nature, whatever in turbulence, life or either markets, tries to maximize the entropy *production*. As Stephen Guerin mentioned earlier this week in one of his lecture, entropy increases but when the system is sufficiently nonlinear, the entropy increases the most rapidly possible. Such overall principle enables to predict the evolution of the system. The book has several chapters showing where his principle could be used: turbulence, mean state of the atmosphere of Earth as well as other planets, the shaping of landscape by water, the coupled evolution of the biosphere and atmosphere, Gaia, economic processes, etc. As long as I do not understand where that principle comes from, I will stay skeptical of those claims. Still, if they are right, this can have important consequences in many fields. Go take a look at the book and if you are interested or if you know something about it, drop me a word. Chuss
Meet to discuss neuroscience problems and, ideally, cutting edge research. I have Kandel's reference text and Arbib's Handbook with me, and will bring them along for the ride and for background. Would also like to speculate on what's going on in your brain and mind :) Maybe we can meet this weekend sometime, or tonight (5/6) at SFI?
Hi Nish, I'm interested (see also my notes above). I could meet early this evening, but otherwise I'll be away for the weekend. Please let me know a time for the evening. Otherwise, I'll set up a tutorial (aiming for a Wednesday AM time next week -- that work for you?) and see if there's interest among other folks. Cheers, Sayres 14:41, 6 June 2008 (MDT)
I'd be interested in throwing my thoughts around. I am not sure if you are going to discuss ideas at a neuronal modeling and analysis level or the behaviour and organization of a network of neurons. srideep
I already went through most of the Kandell textbook during a reading group last year and I would like to discuss more some new stuff like mirror neurons which in that textbook were not present, as well the biochemical aspects of the brain, which we ashamelessly skipped since most of the people in the group were AI folks! Giovanni
Count me in! Chris
Emergence of language in an agent-based model
- I am still interested(Petr).
"Evolving" interactions in M.A.S. (See above my project ideas) Giovanni
Prediction in cultural markets
Time-horizon of political institutions for managing global environmental goods
Social network data/evolution of networks
Networks plus application of biological methods
Public goods gam, group cooperation, social network
...> group behaviour of CSSS 2008
gathering data from all of us, including game theory and perhaps agent based modeling
- Epidemic models of depression/anxiety - Inference of stochastic models
Social-ecological system -> Resilience institutions
Emergence of segregation from a game theory perspective.
1) I have a time series of Chlorophyll a taken over 20 years. A Nature paper has claimed that a chaotic model reproduce the observed time series. Simple question: is there any evidence that the observed time series itself be chaotic? That would be a simple application of the Nonlinear Time Series Analysis introduced by Liz. May just be a nice exercise to do, not necessarily a final project. let me know if interesting in helping me.
2) Lattice-Botlzmann models are quite like the Agent-Based models but they seem better adapted to reproduce fluids in motion: the code is short and I have a couple of them that I got from open sources. I never used them but if anybody is interested to play with it, let me know. we could use the sand table that Redfish has and maybe reproduce some cool stuff such as the flooding of a valley, the breaking of a dam, or something like that!
3) what about the climate? nobody interested? this is one of the most *complex* system. we could try our hands on simple conceptual model of the climate?
Hi, Francois. I'd be interested in helping with (1). Would love to speculate about (3), though I don't have much background. -Sarah
(1) Sounds interesting - The unAustralian Paul
I've had some experience in dealing with problems similar to (1). (3) is an interesting problem - we could play around with the lorenz equation or look at temperature data. I might be interested in in exploring the question as to whether 'global warming' can be reduced or if we were to try and reduce it, can we be certain that we do not over react! This is an interesting problem and I would like to get my feet wet in this sea of ideas. srideep