Transcript Welcome to CSE 590CE: Readings and Research in Computational

Welcome to CSE 590CE:
Readings and Research in
Computational Evolution
Course Mechanics


Mondays 1:30 to 2:20
1/17 and 2/21 are holidays = 8 meetings





Today’s organizational
7 paper discussion meetings
One normal or two small papers per week.
Course web site to be set soon.
Paper presenters should plan on a 30 minute
presentation: 20 slides.
About the Instructor






Daniel Weise
M.S. ’82, PhD ’86 MIT A.I. Lab
Stanford faculty 86-92
Microsoft Research 92-04
Affiliate Faculty (RSN) UW CSE
I’m a CS type learning about biology, cells,
evolution, biochemistry, genetics, ecology,
genomics, proteomics, metabolomics, etc.
We are here to learn and think


We all get to learn together
All comments and insights on papers are
welcome and encouraged


I want this to be a discussion course.
I hope we have a diversity of backgrounds
and approaches in this room to help ensure
we don’t end up in group think
Computational Evolution





It’s about simulation.
Computer power per unit cost is still exploding
exponentially.
Can we use this power to create simulations that
shed insight in biological processes?
What about the compute power available in ten
years?
Instead of post-facto simulations, use compute
power to drive the theory, e.g., Hillis (unpublished)
Computational Evolution:
Self replication + variation + landscapes

Computational models of self-replicating
organisms



Simulated landscapes with niches.


Digital (Von Neumann architecture)
Molecular (communicating processes)
Landscapes provide “fitness” measures
Subject to mutation and variation (diploid)
Building Phenotypes is the Fundamental
Problem in Computational Evolution



Selection operates on the phenotypes of
organisms.
Phenotypes come from physics
Modeling physics is expensive


Approximations
Relating phenotypes back to biology is tricky.
What can we hope to find?





Validation of existing theories/hypotheses.
The ability to propose and test new
hypotheses.
Unanticipated phenomena to look for in
nature (e.g., Hillis)
Better models for the physical world.
Recapitulation of the rise of complexity of
organisms.
CE is at intersection of many fields

Population/Evolutionary Genetics


Ecology


Nature had 10^9 years and 10^28 organisms
Biochemistry and biophysics


When organisms can interact, ecologies form.
Efficient simulation methods


Computes how gene frequencies of populations change
due to selection, migration, & mutation.
When modeling at the molecular level
Artificial Life, Signal Processing, Information Theory,
Program Analysis
Readings









1/10: Evolution, Ecology and Optimization of Digital
Organisms
1/17: Holiday, no class.
1/24: The Evolutionary Origin of Complex Adaptive
Features
1/31: Adaptive Radiation from Resource Competition in
Digital Organisms (2004)
2/7: Evolution of Biological Complexity;
2/14: Tentative: four short Avida papers.
2/21: Holiday, no class.
2/28: TBA
3/07: TBA
Fun Reading






Artificial Life by Steven Levy, Vintage books
Proceedings of the 2nd Artificial Life conf.
Introduction to Artificial Life, Chris Adami,
Telos books
Theoretical Evolutionary Genetics, Joseph
Felsenstein, online at his website
The Philosophy of Artificial Life, Margaret
Boden, Oxford Press
Anything by Dawkins, Gould, or Maynard
Smith