introduction to chaos by: saeed heidary 29 feb 2013
TRANSCRIPT
Introduction to Chaos
by: Saeed Heidary
29 Feb 2013
Outline:Chaos in Deterministic Dynamical systems
Sensitivity to initial conditions
Lyapunov exponent
Fractal geometry
Chaotic time series prediction
Chaos in Deterministic Dynamical systems
There are not any random terms in the equation(s) which describe evolution of the deterministic system.
If the these equations have non-linear term,the
system may be chaotic .
Nonlinearity is a necessary condition but not enough.
Characteristics of chaotic systems
Sensitivity to initial conditions(butterfly effect)
Sensitivity measured by lyapunov exponent.
complex shape in phase space (Fractals ) Fractals are shape with fractional (non integer) dimension !.
Allow short-term prediction but not long-term prediction
Butterfly Effect
The butterfly effect describes the notion that the flapping of the wings of a butterfly can ‘cause’ a typhoon at the other side of the world.
Lyapunov ExponentTow near points in phase space diverge exponentially
Lyapunov exponent
Stochastic (random ) systems:
Chaotic systems :
Regular systems :
0
0
Chaos and RandomnessChaos is NOT randomness though it can look pretty random.
Let us have a look at two time series:
-2
-1.5
-1
-0.5
0
0.5
1
1.5
2
0 50 100 150 200 250 300 350 400-2
-1.5
-1
-0.5
0
0.5
1
1.5
2
0 50 100 150 200 250 300 350 400
Chaos and Randomness
-2
-1.5
-1
-0.5
0
0.5
1
1.5
2
0 50 100 150 200 250 300 350 400
-2
-1.5
-1
-0.5
0
0.5
1
1.5
2
0 50 100 150 200 250 300 350 400
-2
-1.5
-1
-0.5
0
0.5
1
1.5
2
-2 -1.5 -1 -0.5 0 0.5 1 1.5 2
-2
-1.5
-1
-0.5
0
0.5
1
1.5
2
-2 -1.5 -1 -0.5 0 0.5 1 1.5 2
xn+1 = 1.4 - x2n + 0.3 yn
yn+1 = xn
White NoiseNon - deterministic
Henon MapDeterministic
plot xn+1 versus xn (phase space)
fractalsGeometrical objects generally with non-integer
dimensionSelf-similarity (contains infinite copies of itself)Structure on all scales (detail persists when zoomed
arbitrarily)
Fractals productionApplying simple rule against simple shape and iterate it
Fractal production
Sierpinsky carpet
Broccoli fractal!
Box counting dimension
Integer dimensionPoint 0 Line 1
Surface 2
Volume 3
Exercise for non-integer dimensionCalculate box counting dimension for cantor set and
repeat it for sierpinsky carpet?
Fractals in nature
Fractals in nature
Complexity - disorder
Nature is complicated
but
Simple models may suffice
I emphasize:
“Complexity doesn’t mean disorder.”
Prediction in chaotic time seriesConsider a time serie :
The goal is to predict
T is small and in the worth case is equal to inverse of lyapunov exponent of the system (why?)
Forecasting chaotic time series procedure (Local Linear Approximation)The first step is to embed the time series to obtain the
reconstruction (classify)
The next step is to measure the separation distance between the vector and the other reconstructed vectors
And sort them from smalest to largest
The (or ) are ordered with respect to
Local Linear Approximation (LLA) Method
the next step is to map the nearest neighbors of forward forward in the reconstructed phase space for a time T
These evolved points are The components of these vectores are as follows:
Local linear approximation:
Local Linear Approximation (LLA) Method
Again
the unknown coefficients can be solved using a least – squares method
Finally we have prediction
THANK YOU