A Physicist’s Brain J. C. Sprott Department of Physics University of Wisconsin - Madison Presented at the Chaos and Complex Systems Seminar In Madison, Wisconsin On October 18, 2005
Collaborators n David Albers, Max Planck Institute (Leipzig, Germany) n Matt Sieth, Univ Wisc - Undergrad
A Physicist’s Neuron N inputs tanh x x
Architecture N neurons
Artificial Neural Network (P-Brain) n Nonlinear, discrete-time, complex, dynamical system n “Universal” approximator (?) a ij chosen from a random Gaussian distribution with mean zero and standard deviation s Two parameters: N and s Arbitrary (large) N infinity n Initial conditions random in the range -1 to +1.
Probability of Chaos
A Physicist’s EEG
Strange Attractor
Artist’s Brain
Airhead
Dumbbell
Featherbrain
Egghead
Scatterbrain
Attractor Dimension D KY = 0.46 N N
Route to Chaos at Large N (=64)
Animated Route to Chaos
Summary of High- N Dynamics n Chaos is the rule Maximum attractor dimension is of order N /2 n Quasiperiodic route is usual n Attractor is sensitive to parameter perturbations, but dynamics are not
P-Brain Artist n Train a neural network to produce art Choose N = 6 n Find “good” regions of the 36-D parameter space n Randomly explore a neighborhood of that region
Automatic Preselection n Must be chaotic (positive Lyapunov exponent) n Not too “thin” (fractal dimension > 1) n Not too small or too large n Not too off-centered
Training on an Image
Problem – Rugged Landscape Relative Error -5%+5%0
Hurricane Rita
Robin Chapman
Information Content n Robin: 244 x 340 x 3 x 8 = 2 Mbits Compresses (gif) to 283 kbits Compresses (jpeg) to 118 kbits Compresses (png) to 1.8 Mbits n P-Brain: 36 x 5 = 180 bits n Cannot expect a good replica
Future Directions n More biological realism n More neurons n More realistic architecture n Training on real EEG data or task performance
References n lectures/brain.ppt (this talk) lectures/brain.ppt n (contact me)