Blog · Collection

Stochastic differential equations

Ten posts on stochastic calculus: Brownian motion, the Itô integral and Itô's formula, SDEs and their links to PDEs, explosion, martingales, Girsanov's theorem and Lévy processes.

This collection is a tour of stochastic calculus. It starts from measure-theoretic probability, builds Brownian motion and the Itô integral, and uses them to study stochastic differential equations: how to solve them, how they connect to partial differential equations, when their solutions blow up, and how a change of probability measure changes their drift. It ends with processes that jump.

The posts build on each other, so reading them in order works best. They assume probability at the level of random variables and expectations, and some comfort with calculus and real analysis.

Posts in this collection

  1. Probability, the measure-theoretic way 5 min read

    Probability spaces, σ-algebras as information, random variables, modes of convergence and conditional expectation: the language the rest of the collection is written in.

  2. Brownian motion 6 min read

    The definition, its symmetries, three ways to construct it, why its paths have infinite length but finite quadratic variation, and its martingale and Markov properties.

  3. Building the Itô integral 4 min read

    Why the stochastic integral cannot be a Stieltjes integral, how Itô built it from simple adapted processes and an isometry, and what quadratic variation and covariation mean.

  4. Itô's formula 4 min read

    The chain rule of stochastic calculus, from functions of Brownian motion to Itô processes in several dimensions, with worked examples: geometric Brownian motion, exponential martingales, moments and Bessel processes.

  5. Solving stochastic differential equations 5 min read

    What a solution to an SDE is, worked examples (Ornstein–Uhlenbeck, Langevin, the Brownian bridge), the Lipschitz existence and uniqueness theorem, dependence on initial conditions, and what goes wrong without the assumptions.

  6. From SDEs to PDEs 5 min read

    The generator of a diffusion and the single computation behind every link between SDEs and PDEs: backward Kolmogorov equations, Feynman–Kac, boundary value problems and exit times, with Brownian motion on an interval as the running example.

  7. When diffusions explode: Feller's test 4 min read

    Locally Lipschitz SDEs can blow up in finite time. In one dimension, the scale function and speed measure decide exactly when, and noise does not always save the day.

  8. Martingales 5 min read

    Fair games in discrete and continuous time: convergence and uniform integrability, optional stopping, local martingales, maximal inequalities, Lévy's characterization, time changes and the martingale representation theorem.

  9. Girsanov's theorem 4 min read

    Changing the probability measure changes the drift of a Brownian motion: a discrete warm-up, the theorem and why it works, what cannot be changed, and applications from hitting times to finance and rare-event simulation.

  10. Lévy processes and jumps 5 min read

    Processes with independent, stationary increments that can jump: Poisson and compound Poisson processes, Poisson random measures, the Lévy–Khintchine formula, the Lévy–Itô decomposition, and Itô's formula with jumps.

← All collections