• Search |
• Edit this page

Meeting with Santiago

Santiago

These first two I think are not really programming tasks, but math tasks which could be made easier with a script:

• Unit conversions
• Fitting procedures (regression, etc)

These are getting more interesting programatically:

• Matching strings based on multiple relations. For example, taking two strings of codons (I think they are called) and finding matches between them based on various relations, including melting point and things like that.
• Stoichiometric networks. Finding how fast a particular reaction, between two points in such a network, occurs.
• Using network models instead of simple statistical models. For example, simple statistical models of spreading disease aren't very helpful, but if you create a graph of the spread of a disease through individuals, you can do much better analysis.
• Digitizing data. For example, an ecologist collects data about where animals reproduce, which can be plotted on a map over time. How do you get this into structures more advanced than a spreadsheet?
• Clustering. Figuring out in the above situation whether data is clustered, randomly distributed or uniformly distributed.
• Population control. Simulating food, hunting bans, etc, and how they effect population of foxes or something. Marco apparently does ecology simulation
• Resource management. Taxes on garbage, penalties for polution. People like Greenpeace are using linear factors to analyze effects of oils spills and the like, when they are really nonlinear factors.
• 4+ dimensional plots. Both creating the models and doing the visualization.
• Creating lattice models of chemical reactions in space. Using "lattice-gas automata". Right now you have to do a lot of programming, but you should just be able to give the computer an equation A+B->C and have it do the simulation. Plus the computer needs some information about how fast the reactions occur, and you'd like to be able to add some rules about how the thing goes. I would need some more info about that. See Programming lattice models naturally.
• CellML, and other markup languages
• Monte-carlo simulation is a lot like the alttice models, you can use it to look at the physical/chemical properties of water.
• Social insects. Simulating ants forming bridges and whatnot. Emergent behavior.

Categories

I was thinking this would break down into several categories, but in the end it all seems to have a lot to do with creating and measuring models. Simulation seems to be a big common thread.

Sometimes the simulation can be done with discrete models like the lattice model, but I'm not sure the stoichiometric equations really fit. I think there are situations where it is computationally prohivitive/impossible to do a simulation, and you really need to just do out the math.

But Santiago broke this down into three categories:

• equations
• agents
• pattern recognition

Which I think fits with what I said above. Probably we need to just focus on one of these. The bioinformatics students I was talking to seemed to talk mostly about the pattern recognition, whereas Santiago was talking a lot about the first two.