Blog · Collection
Probability theory
Martingales, Markov chains, ergodic theory and Brownian motion: the core of measure-theoretic probability beyond the classical limit theorems.
This overview covers the core of measure-theoretic probability beyond the classical limit theorems, roughly chapters 4 to 7 of Durrett’s Probability: Theory and Examples.
Martingales
A martingale is a fair game: . The main results are the martingale convergence theorem (a martingale that is bounded in converges almost surely), Doob’s maximal inequality, and the optional stopping theorem, which says a fair game stays fair at well-behaved random stopping times. Together they give short proofs of results that are otherwise hard, from the strong law of large numbers to the behavior of branching processes.
Markov chains
For chains on a countable state space, the key distinctions are recurrence versus transience and whether a stationary distribution exists. An irreducible, aperiodic, positive recurrent chain converges to its unique stationary distribution from any starting state. That is the theoretical guarantee behind every MCMC algorithm.
Ergodic theory
The ergodic theorem extends the law of large numbers from independent sequences to stationary ones: averages along a single trajectory converge to averages over the state space. This is exactly why one long MCMC run can estimate expectations.
Brownian motion
Brownian motion ties these topics together. It is a continuous martingale and a strong Markov process, and its paths have a rich structure. The reflection principle gives the distribution of its running maximum, and the law of the iterated logarithm describes how far it strays. It is also the starting point for stochastic calculus.