ebook img

615.18 -- Simulated Annealing PDF

56 Pages·2008·0.42 MB·French
by  
Save to my drive
Quick download
Download
Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.

Preview 615.18 -- Simulated Annealing

SSiimmuullaatteedd AAnnnneeaalliinngg Biostatistics 615/815 LLeeccttuurree 1188 So far … “Greedy” optimization methods (cid:122) • Can get trapped at local minima • Outcome might depend on starting point Examples: (cid:122) • Golden Search • Nelder-Mead Simplex Optimization • E-M Algorithm Today … Simulated Annealing (cid:122) Markov-Chain Monte-Carlo method (cid:122) Designed to search for global minimum (cid:122) aammoonngg mmaannyy llooccaall mmiinniimmaa The Problem Most minimization strategies find the (cid:122) nearest local minimum SSttaannddaarrdd ssttrraatteeggyy (cid:122)(cid:122) • Generate trial point based on current estimates • EEvvaalluuaattee ffuunnccttiioonn aatt pprrooppoosseedd llooccaattiioonn • Accept new value if it improves solution The Solution We need a strategy to find other minima (cid:122) This means, we must sometimes select (cid:122) nneeww ppooiinnttss tthhaatt ddoo nnoott iimmpprroovvee ssoolluuttiioonn HHooww?? (cid:122)(cid:122) Annealing One manner in which crystals are formed (cid:122) Gradual cooling of liquid … (cid:122) • At high temperatures, molecules move freely • At low temperatures, molecules are "stuck" IIff coolliing iis sllow (cid:122) • Low energy, organized crystal lattice formed Simulated Annealing Analogy with thermodynamics (cid:122) Incorporate a temperature parameter into the (cid:122) minimization procedure At high temperatures, explore parameter space (cid:122) At lower temperatures, restrict exploration (cid:122) Markov Chain The Markovian property (cid:122) Pr(Z = i | Z = i , , Z = i ) = Pr(Z = i | Z = i ) K n n n−1 n−1 0 0 n n n−1 n−1 Transition probability (cid:122) QQ == PPrr((ZZ == jj || ZZ == ii)) ij n n−1 n-step transition (cid:122) (n) ∑ ∑ Q = Pr(Z = j | Z = i) = p p L L ij n 0 ii i j 1 n−1 i i 1 n−1 (all possible n-step paths i → j) Markov Chain (In Practice) Start with some state (Z =i) (cid:122) n • A set of mixture parameters Propose a change (Z =j) (cid:122) n+1 •• EEddiitt miixtture parametters iin some way DDeecciiddee wwhheetthheerr ttoo aacccceepptt cchhaannggee ((QQ )) (cid:122)(cid:122) ij • Decision is based on relative probabilities of ttwwoo oouuttccoommeess Simulated Annealing Strategy Consider decreasing series of temperatures (cid:122) For each temperature, iterate these steps: (cid:122) • Propose an update and evaluate function •• AAcceptt upddattes tthhatt iimprove solluttiion • Accept some updates that don't improve solution • Accepptance pprobabilityy deppends on “tempperature” pparameter If cooling is sufficiently slow, the global minimum (cid:122) wiillll bbe reachhedd

Description:
C Code: Simulated Annealing double sa(int k, double * probs, double * means, double * sigmas, double eps) {double llk = -mixLLK(n, data, k, probs, means, sigmas);
See more

The list of books you might like

Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.