Presentation is loading. Please wait.

Presentation is loading. Please wait.

Sampling algorithms and Markov chains László Lovász Microsoft Research One Microsoft Way, Redmond, WA 98052

Similar presentations


Presentation on theme: "Sampling algorithms and Markov chains László Lovász Microsoft Research One Microsoft Way, Redmond, WA 98052"— Presentation transcript:

1 Sampling algorithms and Markov chains László Lovász Microsoft Research One Microsoft Way, Redmond, WA 98052 lovasz@microsoft.com

2 Sampling: a general algorithmic task Applications: - statistics - simulation - counting - numerical integration - optimization - …

3 L: a language in NP, with presentation polynomial time algorithm Find: - a certificate Given: x certificate - an optimal certificate - the number of certificates - a random certificate (uniform, or given distribution)

4 One general method for sampling: Markov chains (+rejection sampling, lifting,…) Construct ergodic Markov chain with states: V stationary distribution: p Want: sample from distribution p on set V Simulate (run) the chain for T steps Output the final state ???????????? mixing time

5 Given: poset 1 2 54 3 State: compatible linear order 1 3 54 2 Transition: - pick randomly label i<n ; - interchange i and i+1 if possible

6 Mixing time Bipartite graph?! : distribution after t steps Roughly: (enough to consider )

7 Conductance in sequence of independent samples: frequency of stepping from S to K\S conductance: in Markov chain:

8 Jerrum - Sinclair But in finer analysis? In typical sampling application: polynomial

9 Key lemma: Proof for l=k+1

10 L – Simonovits Dyer – Frieze Simple isoperimetric inequality: Improved isoperimetric inequality: Kannan-L After appropriate preprocessing,

11 Lifting Markov chains Diaconis – Holmes – Neal


Download ppt "Sampling algorithms and Markov chains László Lovász Microsoft Research One Microsoft Way, Redmond, WA 98052"

Similar presentations


Ads by Google