Lectures
The lectures are given Tue 10:15–12:00 in Room MaD302 and Wed 10:15–12:00 in Room MaD380.
- Tue 1 Nov
- Introduction and course outline.
-
Transition matrix and initial distribution of a Markov chain.
(Häggström Ch2)
- Wed 2 Nov
- Simple random walk on a graph
-
Reachability of states (can a MC starting at state i reach state j?)
(Häggström Ch4)
- Tue 8 Nov
- Simulation of Markov chains: random number generators
(RANDOM.ORG)
- Simulation of Markov chains: random number generators
- Initiation function and update function of a Markov chain
(Häggström Ch3) - Thu 10 Nov (Room MaD353)
- Numerical simulation of Markov chains
(Project assignment 1)
- Numerical simulation of Markov chains
- Tue 15 Nov
- (Ir)reducible Markov chains
(Häggström Ch4) - (A)periodic Markov chains
(Häggström Ch4)
- (Ir)reducible Markov chains
- Wed 16 Nov
-
Strong irreducibility
(Häggström Cor 4.1) - Mean hitting times
(Häggström Lem 5.1)
-
Strong irreducibility
- Tue 22 Nov
- Existence of a stationary distribution
(Häggström Thm 5.1) - Collision Lemma: Independent irreducible aperiodic finite MC's will eventually collide
(Häggström p. 35–36, inside the proof of Thm 5.2)
- Existence of a stationary distribution
- Wed 23 Nov
-
Total variation metric
(Häggström Def 5.2) - Markov Chain convergence theorem
(Häggström Thm 5.2 & Thm 5.3)
-
Total variation metric
- Tue 29 Nov
- Reversible Markov chains
(Häggström Ch 6) - Simple random walk on an undirected graph
(Häggström Ex 6.1) - Metropolis algorithm
(Häggström Ch 7; Levin, Peres & Wilmer Sec 3.2)
- Reversible Markov chains
- Wed 30 Nov
- Modeling optimal scheduling problems using hardcore particle configurations (a.k.a. independent sets in graph theory)
-
A Gibbs sampler (a.k.a. Glauber chain) for generating feasible hardcore configurations
(Häggström Example 7.2) -
Randomized optimization algorithms using low-temperature Boltzmann distributions (a.k.a. Gibbs measures)
(Häggström Theorem 13.1)