Lyapunov exponent quantifies divergence rate: e^(λt)|δ₀| ≈ |δ(t)|
Image: Andrew Magill, CC BY 2.0, via Wikimedia Commons
Lyapunov exponent quantifies divergence rate: e^(λt)|δ₀| ≈ |δ(t)|
The Lyapunov exponent (λ) measures the exponential rate at which trajectories diverge or converge in a dynamical system. This rate of separation is crucial for understanding the system's sensitivity to initial conditions and predicting long-term behavior.
In a dynamical system, two trajectories with an initial separation vector δ₀ diverge at a rate given by e^(λt)|δ₀| ≈ |δ(t)|. The exponent λ indicates how quickly trajectories move apart or come together over time t. Positive λ values suggest divergence, while negative λ values indicate convergence.
Understanding the Lyapunov exponent helps in predicting the stability of a system. Positive Lyapunov exponents are associated with chaotic systems where small differences in initial conditions can lead to vastly different outcomes. Negative exponents, on the other hand, indicate stable systems where trajectories converge to a fixed point or limit cycle.
Example
Consider a simple pendulum with a small initial displacement. If the Lyapunov exponent is positive, the pendulum's motion will diverge exponentially, leading to chaotic behavior. Conversely, if the exponent is negative, the motion will converge, indicating stable oscillations.
Remember this
The Lyapunov exponent provides insight into the predictability and stability of dynamical systems, which is essential for various applications in physics, engineering, and other fields.
Text adapted from Wikipedia, licensed under CC BY-SA 4.0.
Langevin dynamics does: adds noise to gradient descent to sample from a distribution
Langevin dynamics uses stochastic differential equations
Cross-entropy H(p,q) = -Σ p(x) log q(x) measures how well q approximates p
Ever wondered how well we can guess the outcome of a random event?
Chaos theory
Butterfly effect demonstrates sensitive dependence on initial conditions
Curvature
Curvature measures the angular rate of change of the direction of the tangent line per unit distance along the curve
KL divergence is always ≥ 0 and equals 0 only when P = Q exactly
Why can't we just compare two things directly?
the trace equals the sum of eigenvalues: tr(A) = Σλ_i
How can a vector stay the same after a transformation?
Swipe through 100 ML concepts daily
Open Pocket Polymath