21.01.2015 Views

Metropolis-Hastings algorithm

Metropolis-Hastings algorithm

Metropolis-Hastings algorithm

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

General <strong>Metropolis</strong>-<strong>Hastings</strong> <strong>algorithm</strong><br />

• Specific procedure just described was special case:<br />

– Discrete positions (θ);<br />

– One dimension;<br />

– Moves that proposed one position left or right.<br />

• General <strong>algorithm</strong> applies to:<br />

– Continuous values;<br />

– Any number of dimensions;<br />

– More general proposed distributions.<br />

• Essentials are same as for special case:<br />

– Have some target distribution, P(θ), over a multidimensional<br />

continuous parameter space.<br />

– Would like to generate representative samples.<br />

– Must be able to find value of P(θ) for any candidate value of θ.<br />

– Distribution P(θ) does not need to be normalized, merely non-negative.<br />

– Typical Bayesian application: P(θ) is product of likelihood and prior.

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!