
A fixed point is where dx/dt = 0
A fixed point is where dx/dt = 0
A fixed point in a dynamical system is a state where the rate of change is zero, meaning the system remains constant over time. This concept is fundamental in understanding the behavior of systems that evolve according to differential equations.
Example
Consider a simple pendulum. At its equilibrium position, the pendulum momentarily stops before swinging back, representing a fixed point where dx/dt = 0.
Remember this
Identifying fixed points helps predict system behavior and stability, which is crucial for applications in various fields like physics and engineering.
Text adapted from Wikipedia, licensed under CC BY-SA 4.0.
the L1 norm is not differentiable at zero
Absolute value's kink makes it non-differentiable at zero
Physics-informed neural networks
Neural ODEs model continuous-time dynamics with a neural network as the derivative
Tangent space
Tangent space at a point represents all possible velocity vectors
the trace equals the sum of eigenvalues: tr(A) = Σλ_i
How can a vector stay the same after a transformation?
the condition number κ(A) measures: sensitivity of Ax=b to perturbations
Condition number κ(A) measures sensitivity of Ax=b to perturbations
the momentum term does: v_t = βv_{t-1} + ∇L, accumulates gradient direction
Momentum term accelerates convergence in the gradient direction
Swipe through 100 ML concepts daily
Open Pocket Polymath