Ordinary differential equations

An ordinary differential equation (ODE) relates an unknown function to its derivatives with respect to a single variable (often time):

The order is that of the highest derivative. A Cauchy problem (initial value problem) fixes — and possibly derivatives up to order for an ODE of order .

Intuition

An ODE describes dynamic evolution: at each instant, the rate of change of depends on the current state (and possibly time). Knowing is the starting point; solving the ODE traces the trajectory.

Familiar examples in finance and probability: exponential decay (continuous discounting); mean reversion (Ornstein-Uhlenbeck); logistic growth (saturation).

Classification

Type	Form	Comment
Linear		Standard analytical methods (integrating factor)
Homogeneous	with linear in	Superposition of solutions
Autonomous	(no explicit )	Stationary vector field
Separable		Direct integration after separation
Nonlinear	E.g. logistic, Lotka-Volterra	Often no closed form; numerical solution
System		Coupling between several state variables

Classic examples

ODE	Solution	Interpretation
		Exponential growth/decay
		Mean reversion toward
	Implicit solution;	Logistic, saturation at
		Harmonic oscillator
(no noise)	Deterministic backbone of GBM	Prelude to stochastic processes

Solution methods

Separation of variables when . - Integrating factor for first-order linear ODEs. - Characteristic equation for constant-coefficient linear ODEs of order . - Taylor / power series for formal solutions — see Taylor series.

When no closed formula exists:

Runge-Kutta (RK4) and implicit schemes improve stability and accuracy. See Numerical methods for finite differences and scheme stability.

Key theorems

Existence and uniqueness (Cauchy-Lipschitz): if is continuous and Lipschitz in , the Cauchy problem has a unique local solution.

Consequences: no ambiguity in the trajectory for a given initial condition; sensitivity to parameters is controlled (variational equations); numerical simulators behave predictably if is small enough.

Applications

Finance: rate models (Vasicek, CIR mean dynamics), continuous discounting, noise-free fund dynamics.
Probability: deterministic ODEs for process moments; bridge to SDEs when a diffusion term is added.
Physics / engineering: circuits, mechanics, optimal control (Pontryagin's maximum principle).
Machine learning: Neural ODEs (networks defined as ODEs) — same formalism, numerical backprop through the solver.

PDEs generalize the formalism to several independent variables (time + space, or time + market factors).

Limits and pitfalls

Situation	Issue	Consequence
Non-Lipschitz	E.g.	Uniqueness may fail (multiple trajectories from 0)
Stiffness	Very different time scales	Explicit Euler unstable; implicit schemes needed
Chaos	Extreme sensitivity to I.C.	Long-term prediction impossible despite determinism
Confusing ODE and SDE	GBM is not an ODE	Stochastic calculus (Itô) required
Step too large	Discretization error	Wrong or divergent numerical solution