Actions

Complex Systems Summer School 2017-Projects & Working Groups

From Santa Fe Institute Events Wiki

Revision as of 13:47, 13 June 2017 by Bleu (talk | contribs) (→‎Resources)
Complex Systems Summer School 2017

suggestions from Jiří Moravec

Microbial Systems are Complex

Summary

Several of us seem to have shared interests in microbial ecology/evolution/genetics. It might be a good idea to meet up and brainstorm a project (maybe over beer?). Some potential ideas that were discussed during speed dating:

  • The evolution of the bacterial pangenome
  • Development/adaptation of classical population genetic techniques to microbes (maybe via a network describing HGT interactions)
  • The evolution of metabolism/the interaction between metabolism and community structure
  • How interspecific interactions (e.g. crossfeeding, competition) influence the rate of decay of community similarity in a homogeneous environment (scale of spatial organization)

Interested Participants

  • Jake Weissman (jw4336(at)umd.edu)


EU: responding to migrant crisis

Keywords

cooperation, migration, consensus-reaching, supranational unions

Summary

How does an organization of independent countries, such the European Union, deal with a migrant crisis? How can a consensus on a response be reached? Why did the EU fail to deal efficiently and fairly with the 2015 migrant crisis (>1m people)? Why isn’t the migrant quota system not working? (Or is it?)

Yes, these are all pretty big questions, but I believe very exciting and timely. Haven’t figured out details obviously, but I think these could be explored with agent based models (possibly NetLogo?). It would be particularly interesting to see how a few factors influence the chances that a consensus is reached and what type of consensus:

  • country on coast/external border vs landlocked
  • how far from coastal border
  • economic power
  • diplomatic power
  • Union organization: continuum - federal government can impose policies —> unanimous consensus must be reached every time

If you find this interesting, include your name and email on the list below & let’s meet up and discuss!

Resources

Interested Participants

  • Alberto Micheletti (biology, social evolution, ajcm2[at]st-andrews[dot]ac[dot]uk) - group contact
  • Andrew Johnson (biology, ecology, competition, collapse, afjohnson[at]ucsd[dot]edu)
  • Ella?
  • Madison?

Gut-Brain Axis: Impact of gut microbiota on mental health

Keywords

public health, microbiome, metabolism, network analysis, neuroscience

Summary

The term "gut feeling" is not without a scientific basis. Recent literature has emphasized the connection between the brain and intestinal microbes. Studies are beginning to link neurodegenerative disorders, such as Alzheimer's disease and multiple sclerosis, to gut dysbioses.

Other relationships between the brain and the gut are waiting to be explored. Also, on a broader level, relationships between microbial metabolism and neurotransmitter levels could be investigated.

Open source data sets are available, such as the ones mentioned in this paper: bmcsystbiol.biomedcentral.com/articles/10.1186/s12918-016-0307-y

If you find this interesting, include your name and email on the list below

Interested Participants

Bleu Knight (bleu AT nmsu.edu; happylittleaccident AT gmail.com)

Beer: evolution through the ages

Keywords

beer, evolution, genetic elements, history

Summary

Brewing beer is an ancient art that has experienced a renaissance in the last few decades. Can we model beer recipes in an "evolutionary" sense? Can ingredients like hops and malt be modeled like genes that are expressed more in some beers than in others? Can we create phylogenetic trees that delineate relationships between beers? Can beers be considered 'species', and if so, what has contributed to the diversity that we see in the US today?

If you find this interesting, include your name and email on the list below

Media:5000_yr_old_Chinese_beer_recipe.pdf===Resources=== Media:Early_Iron_Age_and_Late_Mediaeval_malt.pdf

Interested Participants

Bleu Knight (bleu AT nmsu.edu; happylittleaccident AT gmail.com)

Agent-Based Prediction Competition

Keywords

Agent-Based Models, Machine Learning, Prediction

Summary

Agent-Based Models (ABMs) have so far been used mainly for descriptive purposes. Some people (including myself) think that ABMs can unleash their full potential in making predictions. I envision this project as an open-ended test to this hypothesis. The main idea is borrowed from the machine learning prediction competitions (www.kaggle.com), in which you are given a training dataset and have to outperform the other participants in predicting the outputs of a test set. Instead of using a machine learning algorithm to approximate a target function, I propose we use an ABM. This project can take several directions:

  • The topic of the ABM and the data can come from any discipline: biology, epidemiology, economics, sociology, ecology, etc.
  • The data can be real or synthetic. In the latter case, someone could code a very complicated ABM as the Data Generating Process of a fictitious system, and the task would be to approximate this complicated ABM.
  • A humble goal would be to take machine learning as an upper bound for prediction possibilities (https://site.stanford.edu/sites/default/files/submission_kleinbergliangmullainathan.pdf), and see how putting realistic assumptions in the ABM makes the prediction score close to the machine learning benchmark.
  • A more ambitious goal would be to beat the machine learning algorithms. We know they can fail because the world is non-stationary (see Google Flu Trends failure). ABMs can account for that.
  • Any other direction!

I would be happy to discuss about any practical example with whoever is interested. I think this project can be very risky, but this is why I’m proposing it here! Still, at least a proof of principle would be a great result. I'm open to any hijacking of this project however!

Group Contact

Marco Pangallo (complexity economics, marco[dot]pangallo[at]maths[dot]ox[dot]ac[dot]uk)

Interested Participants

AI-aided Graph Auto-generation

Keywords

Graph Auto-generation; Graph Characteristics; Artificial Intelligence; Combinatorial Optimization;

Summary

Complex networks/graphs appear in almost every aspect of science and technology, ranging from social systems, computer networks, biological networks to the state spaces of physical systems, even Bayesian Belief Networks. Graph auto-generation has wide applications. Just like what is done by the Google AlphaGo, this oftentimes ends up as a combinatorial optimization problem. The prohibitive number of possibilities calls upon the Artificial Intelligence to determine better strategic choices at every turn. There are several interesting topics.

  • As a starter, we would like to determine the topological structure of given networked systems. A variety of basic measures and metrics are available that can tell us about small-scale structure in networks, such as correlations, connections and recurrent patterns, but it is considerably more difficult to quantify structure on medium and large scales which can help us understand the ‘big picture’.
  • Next, a fundamental property of interdependent networks is that failure of nodes in one network may lead to failure of dependent nodes in other networks. This may happen recursively and can lead to a cascade of failures. In fact, a failure of a very small fraction of nodes in one network may lead to the complete fragmentation of a system of several interdependent networks.
  • Thirdly, assume we have a collection of node and a defined dependency model. Can we auto-assemble these node to form a network which satisfies performance requirements. The dependency model can be defined in the following way
   1. Each node has a defined interface, which specifies what kind of supplies it needs and what kind output it can produce.
   2. The output of a given node can be computed based on its chosen suppliers and available supply levels.
  • Finally, given a set of nodes, assuming the edge length between any pair of nodes is computable, can we find a shortest path between two given nodes such that the path consists of edges with bounded length. Moreover, given multiple measures of the edge length, can we find a Pareto front of shortest paths? In other words, a multi-objective optimization of such paths.

Group Contact

Huang Tang (MITRE, Co., htang@mitre.org)

Interested Participants


Some datasets that might trigger an idea

DataUSA - bunch of USA data (towns, cities, states, education, skills, occupations, industries...)

https://datausa.io

Pantheon - historical database of 'globally known people' with place and date of birth and death (not sure if it's publicly available though..)

http://pantheon.media.mit.edu


Ethnographic Atlas

Retrieved from "https://wiki.santafe.edu/index.php?title=Complex_Systems_Summer_School_2017-Projects_%26_Working_Groups&oldid=67014"