11.07.2015 Views

Network determination based on birth-death MCMC inference

Network determination based on birth-death MCMC inference

Network determination based on birth-death MCMC inference

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

NETWORK BIRTH-DEATH <strong>MCMC</strong> METHOD SPECIFIC ELEMENT OF BD<strong>MCMC</strong> METHOD EXAMPLES<str<strong>on</strong>g>Network</str<strong>on</strong>g> <str<strong>on</strong>g>determinati<strong>on</strong></str<strong>on</strong>g> <str<strong>on</strong>g>based</str<strong>on</strong>g> <strong>on</strong> <strong>birth</strong>-<strong>death</strong><strong>MCMC</strong> <strong>inference</strong>A. Mohammadi and E. WitFebruary 4, 2013


NETWORK BIRTH-DEATH <strong>MCMC</strong> METHOD SPECIFIC ELEMENT OF BD<strong>MCMC</strong> METHOD EXAMPLESWHAT IS TALK ABOUT ?Problem◮ High-dimensi<strong>on</strong>al cases: p(p − 1)/2 ≫ n◮ Bayesian approches : Not fast◮ glasso : Sensitivity to tuning parametersSoluti<strong>on</strong>◮ We proposed Bayesian method which is fast and accurate◮ Implement to R package: BDgraphTrans-dimensi<strong>on</strong>al <strong>MCMC</strong>◮ Reversible-jump <strong>MCMC</strong>◮ Birth-Death <strong>MCMC</strong>


NETWORK BIRTH-DEATH <strong>MCMC</strong> METHOD SPECIFIC ELEMENT OF BD<strong>MCMC</strong> METHOD EXAMPLES<str<strong>on</strong>g>Network</str<strong>on</strong>g>Gaussian graphical model with respect to graph G = (V, E) asM G ={}N p (0, Σ) | K = Σ −1 is positive definite <str<strong>on</strong>g>based</str<strong>on</strong>g> <strong>on</strong> GPairwise Markov propertyX i ⊥X j | X V\{i,j} ⇔ k ij = 0,


NETWORK BIRTH-DEATH <strong>MCMC</strong> METHOD SPECIFIC ELEMENT OF BD<strong>MCMC</strong> METHOD EXAMPLESBIRTH-DEATH PROCESS◮ Spacial <strong>birth</strong>-<strong>death</strong> process: Prest<strong>on</strong> (1976)◮ Brith-<strong>death</strong> <strong>MCMC</strong>: Stephen (2000) in mixture models


NETWORK BIRTH-DEATH <strong>MCMC</strong> METHOD SPECIFIC ELEMENT OF BD<strong>MCMC</strong> METHOD EXAMPLESBIRTH-DEATH <strong>MCMC</strong> DESIGNGeneral <strong>birth</strong>-<strong>death</strong> process◮ C<strong>on</strong>tinuous Markov process◮ Birth and <strong>death</strong> events are independent Poiss<strong>on</strong> processes◮ Time of <strong>birth</strong> or <strong>death</strong> event is exp<strong>on</strong>entially distributedBirth-<strong>death</strong> process in GGM◮ Adding new edge in <strong>birth</strong> time and deleting edge in <strong>death</strong>time


NETWORK BIRTH-DEATH <strong>MCMC</strong> METHOD SPECIFIC ELEMENT OF BD<strong>MCMC</strong> METHOD EXAMPLESSIMPLE CASE


NETWORK BIRTH-DEATH <strong>MCMC</strong> METHOD SPECIFIC ELEMENT OF BD<strong>MCMC</strong> METHOD EXAMPLESSIMPLE CASE


NETWORK BIRTH-DEATH <strong>MCMC</strong> METHOD SPECIFIC ELEMENT OF BD<strong>MCMC</strong> METHOD EXAMPLESSIMPLE CASE


NETWORK BIRTH-DEATH <strong>MCMC</strong> METHOD SPECIFIC ELEMENT OF BD<strong>MCMC</strong> METHOD EXAMPLESSIMPLE CASE


NETWORK BIRTH-DEATH <strong>MCMC</strong> METHOD SPECIFIC ELEMENT OF BD<strong>MCMC</strong> METHOD EXAMPLESSIMPLE CASE


NETWORK BIRTH-DEATH <strong>MCMC</strong> METHOD SPECIFIC ELEMENT OF BD<strong>MCMC</strong> METHOD EXAMPLESSIMPLE CASE


NETWORK BIRTH-DEATH <strong>MCMC</strong> METHOD SPECIFIC ELEMENT OF BD<strong>MCMC</strong> METHOD EXAMPLESBALANCE CONDITIONPrest<strong>on</strong> (1976): Backward KolmogorovIf balance c<strong>on</strong>diti<strong>on</strong>s are hold, process c<strong>on</strong>verges to uniquestati<strong>on</strong>ary distributi<strong>on</strong>.


NETWORK BIRTH-DEATH <strong>MCMC</strong> METHOD SPECIFIC ELEMENT OF BD<strong>MCMC</strong> METHOD EXAMPLESPROPOSED BIRTH-DEATH <strong>MCMC</strong> ALGORITHMProposal <strong>birth</strong> and <strong>death</strong> ratesβ ξ (K) = λ b , new link ξ = (i, j)δ ξ (K) = b ξ(k ξ )p(G − ξ , K− ξ |x)λ b , existing link ξ = (i, j)p(G, K|x)Proposal <strong>birth</strong>-<strong>death</strong> <strong>MCMC</strong> algorithmStarting with initial graph:Step 1: (a). Calculate <strong>birth</strong> and <strong>death</strong> rates(b). Calculate waiting time, λ(K) = 1/(β(K) + δ(K))(c). Simulate type of jump, <strong>birth</strong> or <strong>death</strong>Step 2: Sampling from new precisi<strong>on</strong> matrix: K + ξor K− ξ


NETWORK BIRTH-DEATH <strong>MCMC</strong> METHOD SPECIFIC ELEMENT OF BD<strong>MCMC</strong> METHOD EXAMPLESSAMPLING BIRTH-DEATH <strong>MCMC</strong> ALGORITHM


NETWORK BIRTH-DEATH <strong>MCMC</strong> METHOD SPECIFIC ELEMENT OF BD<strong>MCMC</strong> METHOD EXAMPLES◮ Birth-<strong>death</strong> <strong>MCMC</strong> algorithm for general case ...


NETWORK BIRTH-DEATH <strong>MCMC</strong> METHOD SPECIFIC ELEMENT OF BD<strong>MCMC</strong> METHOD EXAMPLESSPECIFIC ELEMENT OF BD<strong>MCMC</strong> METHOD◮ Prior distributi<strong>on</strong>s◮ Computing <strong>death</strong> rates◮ Sampling from precisi<strong>on</strong> matrix


NETWORK BIRTH-DEATH <strong>MCMC</strong> METHOD SPECIFIC ELEMENT OF BD<strong>MCMC</strong> METHOD EXAMPLESPROPOSED PRIOR DISTRIBUTIONSPrior for graph◮ Discrete Uniform◮ Truncated Poiss<strong>on</strong> according to number of linksPrior for precisi<strong>on</strong> matrix◮ G-Wishart: W G (b, D)p(K|G) ∝ |K| (b−2)/2 exp{− 1 }2 tr(DK)∫I G (b, D) = |K| (b−2)/2 exp{− 1 }P G2 tr(DK) dK


NETWORK BIRTH-DEATH <strong>MCMC</strong> METHOD SPECIFIC ELEMENT OF BD<strong>MCMC</strong> METHOD EXAMPLESG-WISHART DISTRIBUTIONSampling from G-Wishart distributi<strong>on</strong>◮ Accept-reject algorithm◮ Metropolis-Hastings algorithm◮ Block Gibbs sampler◮ According to maximum cliques◮ Edgewise block Gibbs sampler


NETWORK BIRTH-DEATH <strong>MCMC</strong> METHOD SPECIFIC ELEMENT OF BD<strong>MCMC</strong> METHOD EXAMPLESCOMPUTING DEATH RATESδ ξ (K) = p(G− ξ , K− ξ |x)p(G, K|x) γ bb ξ (k ξ )=I G (b, D)(|K −ξ |(b, D) |K|I G−ξ) (b ∗ −2)/2 {exp − 1 }2 tr(D∗ (K −ξ − K)) γ b b ξLimitati<strong>on</strong>Algorithm is very slow !!


NETWORK BIRTH-DEATH <strong>MCMC</strong> METHOD SPECIFIC ELEMENT OF BD<strong>MCMC</strong> METHOD EXAMPLESRatio of normalizing c<strong>on</strong>stantI G (b, D)I G −ξ(b, D) = 2√ Γπt ii t jjΓ( ) b+νi2( b+νi −12) E G [f T (ψ ν )]E G −ξ [f T (ψ ν )]Plot for ratio of normalizing c<strong>on</strong>stantsratio of expectati<strong>on</strong>1.00 1.05 1.10 1.15 1.200 50 100 150 200 250 300number of nodes


NETWORK BIRTH-DEATH <strong>MCMC</strong> METHOD SPECIFIC ELEMENT OF BD<strong>MCMC</strong> METHOD EXAMPLESDeath rates for high-dimensi<strong>on</strong>al casesδ ξ (K) = 2 √ πt ii t jjΓ( ) b+νi2( ) b+νi −12Γ{× exp − 1 2 tr(D∗ (K −ξ − K))) (b ∗ −2)/2(|K −ξ ||K|}γ b b ξ (k ξ )R packageWe compile our method into BDgraph package which isavailable from CRAN web site


NETWORK BIRTH-DEATH <strong>MCMC</strong> METHOD SPECIFIC ELEMENT OF BD<strong>MCMC</strong> METHOD EXAMPLESSIMULATION: 8 NODES}M G ={N 8 (0, Σ)|K = Σ −1 ∈ P G⎡⎤1 .5 0 0 0 0 0 .41 .5 0 0 0 0 01 .5 0 0 0 0K =1 .5 0 0 01 .5 0 01 .5 0⎢⎥⎣1 .5⎦1


NETWORK BIRTH-DEATH <strong>MCMC</strong> METHOD SPECIFIC ELEMENT OF BD<strong>MCMC</strong> METHOD EXAMPLESSOME RESULTEffect of Sample sizeNumber of data 20 30 40 60 80 100 150p(true graph | data) 0.018 0.067 0.121 0.2 0.22 0.35 0.43false discovery 1 0 0 0 0 0 0false negative 0 0 0 0 0 0 0


NETWORK BIRTH-DEATH <strong>MCMC</strong> METHOD SPECIFIC ELEMENT OF BD<strong>MCMC</strong> METHOD EXAMPLESSOME RESULT⎡⎤1 1 0.03 0.06 0.02 0.02 0.03 11 1 0.04 0.03 0.02 0.03 0.031 1 0.06 0.04 0.06 0.03ˆp ξ =1 1 0.05 0.04 0.031 1 0.05 0.13.1 1 0.13⎢⎥⎣1 1 ⎦1⎡⎤1.3 0.6 0 0 0 0 0 0.51.4 0.5 0 0 0 0 01 0.5 0 0 0 0ˆK =1.2 0.6 0 0 01.3 0.4 0 0.0.9 0.5 0⎢⎥⎣0.9 0.4⎦


NETWORK BIRTH-DEATH <strong>MCMC</strong> METHOD SPECIFIC ELEMENT OF BD<strong>MCMC</strong> METHOD EXAMPLESSIMULATION: 120 NODESM G ={N 120 (0, Σ)|K = Σ −1 ∈ P G},◮ n = 1000 ≪ 7260◮ Priors: K ∼ W G (3, I 120 ) and G ∼ TU(all possible graphs)◮ 10000 iterati<strong>on</strong>s and 5000 iterati<strong>on</strong>s as burn-inResult◮ Time 190 minutes◮ p(true graph | data) = 0.41 which is most probable graph


NETWORK BIRTH-DEATH <strong>MCMC</strong> METHOD SPECIFIC ELEMENT OF BD<strong>MCMC</strong> METHOD EXAMPLESCELL SIGNALING DATAFlow cytometry data with 11 proteins from Sachs et al. (2005)(Left) Result from our algorithm(Right) Result from Sachs et al (2005)Friedman et al (2008): full graph according to g-lasso


NETWORK BIRTH-DEATH <strong>MCMC</strong> METHOD SPECIFIC ELEMENT OF BD<strong>MCMC</strong> METHOD EXAMPLESCONCLUSION


NETWORK BIRTH-DEATH <strong>MCMC</strong> METHOD SPECIFIC ELEMENT OF BD<strong>MCMC</strong> METHOD EXAMPLESThanks for your attenti<strong>on</strong>ReferencesMOHAMMADI, A. AND E. C. WIT (2012) Efficient <strong>birth</strong>-<strong>death</strong> <strong>MCMC</strong> <strong>inference</strong> forGaussian graphical models. arXiv preprint arXiv:1210.5371WANG, H. AND S. LI (2012) Efficient Gaussian graphical model <str<strong>on</strong>g>determinati<strong>on</strong></str<strong>on</strong>g> underG-Wishart prior distributi<strong>on</strong>s. Electr<strong>on</strong>ic Journal of Statistics, 6:168-198ATAY-KAYIS, A. AND H. MASSAM (2005) A M<strong>on</strong>te Carlo method for computing themarginal likelihood in n<strong>on</strong>decomposable Gaussian graphical models. Biometrika Trust,92(2):317-335PRESTON, C. J. (1976) Special <strong>birth</strong>-and-<strong>death</strong> processes. Bull. Inst. Internat. Statist.,34:1436-1462

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

Saved successfully!

Ooh no, something went wrong!