Assessing the Goodness-of-Fit of Network Models
|
|
- Willa Bradford
- 5 years ago
- Views:
Transcription
1 Assessing the Goodness-of-Fit of Network Models Mark S. Handcock Department of Statistics University of Washington Joint work with David Hunter Steve Goodreau Martina Morris and the U. Washington Network Modeling Group Supported by NIH NIDA Grant DA023 and NICHD Grant HD04 NIPS Networks Workshop 200, December 200
2 Examples of Friendship Relationships The National Longitudinal Study of Adolescent Health Add Health" is a school-based study of the health-related behaviors of adolescents in grades to 2. Each nominated up to 5 boys and 5 girls as their friends 60 schools: Smallest has 6 adolescents in grades 2
3
4 Grade Grade Grade Grade 0 Grade Grade 2 White (non-hispanic) Black (non-hispanic) Hispanic (of any race) Asian / Native Am / Other (non-hispanic) Race NA
5 Statistical Models for Social Networks Notation A social network is defined as a set of n social actors" and a social relationship between each pair of actors.
6 Statistical Models for Social Networks Notation A social network is defined as a set of n social actors" and a social relationship between each pair of actors. { relationship from actor i to actor j Y ij = 0 otherwise
7 Statistical Models for Social Networks Notation A social network is defined as a set of n social actors" and a social relationship between each pair of actors. { relationship from actor i to actor j Y ij = 0 otherwise call Y [Y ij ] n n a sociomatrix a N = n(n ) binary array
8 Statistical Models for Social Networks Notation A social network is defined as a set of n social actors" and a social relationship between each pair of actors. { relationship from actor i to actor j Y ij = 0 otherwise call Y [Y ij ] n n a sociomatrix a N = n(n ) binary array The basic problem of stochastic modeling is to specify a distribution for Y i.e., P(Y = y)
9 A Framework for Network Modeling Let Y be the sample space of Y e.g. {0, } N Any model-class for the multivariate distribution of Y can be parametrized in the form: P η (Y = y) = exp{η g(y)} κ(η, Y) y Y η Λ R q q-vector of parameters Besag (4), Frank and Strauss (6) g(y) q-vector of network statistics. g(y ) are jointly sufficient for the model For a saturated" model-class q = 2 Y κ(η, Y) distribution normalizing constant κ(η, Y) = y Y exp{η g(y)}
10 Approximating the loglikelihood Suppose Y, Y 2,..., Y m i.i.d. P η0 (Y = y) for some η 0. Using the LOLN, the difference in log-likelihoods is l(η) l(η 0 ) = log κ(η 0) κ(η)
11 Approximating the loglikelihood Suppose Y, Y 2,..., Y m i.i.d. P η0 (Y = y) for some η 0. Using the LOLN, the difference in log-likelihoods is l(η) l(η 0 ) = log κ(η 0) κ(η) = log E η0 (exp {(η 0 η) g(y )})
12 Approximating the loglikelihood Suppose Y, Y 2,..., Y m i.i.d. P η0 (Y = y) for some η 0. Using the LOLN, the difference in log-likelihoods is l(η) l(η 0 ) = log κ(η 0) κ(η) = log E η0 (exp {(η 0 η) g(y )}) log M exp {(η 0 η) (g(y i ) g(y obs ))} M i= l(η) l(η 0 ).
13 Approximating the loglikelihood Suppose Y, Y 2,..., Y m i.i.d. P η0 (Y = y) for some η 0. Using the LOLN, the difference in log-likelihoods is l(η) l(η 0 ) = log κ(η 0) κ(η) = log E η0 (exp {(η 0 η) g(y )}) log M exp {(η 0 η) (g(y i ) g(y obs ))} M i= l(η) l(η 0 ). Simulate Y, Y 2,..., Y m using a MCMC (Metropolis-Hastings) algorithm Handcock (2002).
14 Approximating the loglikelihood Suppose Y, Y 2,..., Y m i.i.d. P η0 (Y = y) for some η 0. Using the LOLN, the difference in log-likelihoods is l(η) l(η 0 ) = log κ(η 0) κ(η) = log E η0 (exp {(η 0 η) g(y )}) log M exp {(η 0 η) (g(y i ) g(y obs ))} M i= l(η) l(η 0 ). Simulate Y, Y 2,..., Y m using a MCMC (Metropolis-Hastings) algorithm Handcock (2002). Approximate the MLE ˆη = argmax η { l(η) l(η 0 )} (MC-MLE) Geyer and Thompson (2)
15 Estimating the change in log-likelihood Theoretically, the estimated value of l(θ) l(θ 0 ) converges to the true value as the size of the MCMC sample increases, regardless of the value of θ 0.
16 Estimating the change in log-likelihood Theoretically, the estimated value of l(θ) l(θ 0 ) converges to the true value as the size of the MCMC sample increases, regardless of the value of θ 0. However, in practice this convergence can be agonizingly slow, especially if θ 0 is not chosen close to the maximizer of the likelihood. Hunter and Handcock (2006)
17 Measures of the Goodness-of-fit of models As this is an Exponential family, natural to measure goodness-of-fit via deviance [ ] deviance = 2 l(saturated model) l(ˆθ)) and residual deviance = 2 [ ] l(ˆθ) l(0)
18 Measures of the Goodness-of-fit of models As this is an Exponential family, natural to measure goodness-of-fit via deviance [ ] deviance = 2 l(saturated model) l(ˆθ)) and residual deviance = 2 [ ] l(ˆθ) l(0) Standard" asymptotic arguments approximate this by a χ 2 distribution
19 Measures of the Goodness-of-fit of models As this is an Exponential family, natural to measure goodness-of-fit via deviance [ ] deviance = 2 l(saturated model) l(ˆθ)) and residual deviance = 2 [ ] l(ˆθ) l(0) Standard" asymptotic arguments approximate this by a χ 2 distribution The standard asymptotic approximation can be very bad here...
20 Measures of the Goodness-of-fit of models As this is an Exponential family, natural to measure goodness-of-fit via deviance [ ] deviance = 2 l(saturated model) l(ˆθ)) and residual deviance = 2 [ ] l(ˆθ) l(0) Standard" asymptotic arguments approximate this by a χ 2 distribution The standard asymptotic approximation can be very bad here... but the deviance may still be a useful measure of fit if properly calibrated. Hunter and Handcock (2006)
21 How can we tell if a model class is useful? Many aspects: Is the model-class itself able to represent a range of realistic networks? model degeneracy: small range of graphs covered as the parameters vary (Handcock 2003)
22 How can we tell if a model class is useful? Many aspects: Is the model-class itself able to represent a range of realistic networks? model degeneracy: small range of graphs covered as the parameters vary (Handcock 2003) What are the properties of different methods of estimation? e.g, MLE, psuedolikelihood, Bayesian framework computational failure: estimates do not exist for certain observable graphs
23 How can we tell if a model class is useful? Many aspects: Is the model-class itself able to represent a range of realistic networks? model degeneracy: small range of graphs covered as the parameters vary (Handcock 2003) What are the properties of different methods of estimation? e.g, MLE, psuedolikelihood, Bayesian framework computational failure: estimates do not exist for certain observable graphs Can we assess the goodness-of-fit of models? appropriate measures and tests (Besag 2000; Hunter, Goodreau, Handcock 200)
24 Model Degeneracy idea: A random graph model is near degenerate if the model places almost all its probability mass on a small number of graph configurations in Y. e.g. empty graph, full graph, an individual graph, no 2 stars, mono-degree graphs
25 Model Degeneracy idea: A random graph model is near degenerate if the model places almost all its probability mass on a small number of graph configurations in Y. e.g. empty Classes graph, of statistics full graph, usedan forindividual modeling graph, no 2 stars, mono-degree graphs ) Nodal Markov statistics Frank and Strauss (6) Example: The 2-star model motivated by notions of symmetry and homogeneity P(Y = y) = exp{η edges in Y that do note(y) share an+ actor η 2 S(y)} are c(η, η 2 ) is near-degenerate for most values of η 2 > 0 E(y) = y ij S(y) = y ij y ik i<j i<j<k y Y conditionally independent given the rest of the network analogous to nearest neighbor ideas in spatial statistics Degree distribution: d k (y) = proportion of nodes of degree k in y. k-star distribution: s k (y) = proportion of k-stars in the graph y. triangles: t (y) = proportion of triangles in the graph y.. j 3 h i i. i j j j 2 j j 2 triangle two-star three-star = transitive triad Mark S. Handcock Statistical Modeling With ERGM Figure: Some configurations for non-directed graphs
26
27
28 Geometry of Exponential Random Graph Models Consider the alternative parametrization of the models µ : Λ int(c) defined by µ(η) = E η [Z (Y )] y Y Z (y) exp{ηt Z (y)} c(η) The mapping is injective: µ(η a ) = µ(η b ) P ηa (Y = y) = P ηb (Y = y) y. The mapping in strictly increasing in the sense that (η a η b ) T (µ(η a ) µ(η b )) 0 with equality only if P ηa (Y = y) = P ηb (Y = y) y. Represents an alternative parameterization of the model
29 Example of the 2 star model P(Y = y) = exp{η E(y) + η 2 S(y)} y Y c(η, η 2 ) where E(y) is the number of edges (0 N = ( g 2) ) S(y) is the number of 2 stars (0 M = 3 ( g 3) )
30 Example of the 2 star model P(Y = y) = exp{η E(y) + η 2 S(y)} y Y c(η, η 2 ) where E(y) is the number of edges (0 N = ( g 2) ) S(y) is the number of 2 stars (0 M = 3 ( g 3) ) µ = E η [E(Y )] = i<j E[Y ij ] = NE[Y 2 ] µ is the expected number of edges, or N µ is the probability that two actors are linked.
31 Example of the 2 star model P(Y = y) = exp{η E(y) + η 2 S(y)} y Y c(η, η 2 ) where E(y) is the number of edges (0 N = ( g 2) ) S(y) is the number of 2 stars (0 M = 3 ( g 3) ) µ = E η [E(Y )] = i<j E[Y ij ] = NE[Y 2 ] µ is the expected number of edges, or N µ is the probability that two actors are linked. µ 2 = E η [S(Y )] = E[Y ij Y ik ] = ME[Y 2 Y 3 ] i<j<k µ 2 is the expected number of 2 stars, or M µ 2 is the probability that a given actor is tied to two randomly chosen other actors.
32 Figure 4: Regions of the parameter space of µ µ 2 : mean 2 stars parameter µ : mean edges parameter
33 Figure 5: Regions of the parameter space of θ θ 2 : 2 stars parameter θ : edges parameter
34 In some cases mixed parameterizations may be better Let (t (), t (2) ) be a partition of t such that: t () is interpretable as a mean value parametrization t (2) is interpretable as the natural" conditional log-odds Consider similar partitions (η (), η (2) ) of η and (µ () (η), µ (2) (η)) of µ(η). Let Λ (2) be the set of values of η (2) for η varying in Λ and C () be the convex hull of {t () (y) : y Y}. The mapping η : Λ Λ (2) int(c () ) defined by η(η) = (µ () (η), η (2) ) () is a mixed parametization of the model (Y, t, η). The components µ () and η (2) are variationally independent, that is, the range of η(η) is a product space.
35 Degeneracy in the mean value parametization Definition: A model is near degenerate if µ(η) is close to the boundary of C
36 Degeneracy in the mean value parametization Definition: A model is near degenerate if µ(η) is close to the boundary of C Let deg Y = {y Y : Z (y) bdc} be the set of graph on the boundary of the convex hull. idea: Based on the geometry of the mean value parametrization the expected sufficient statistics are close to a boundary of the hull and the model will place much probability mass on graphs in deg Y.
37 Figure 4: Regions of the parameter space of µ µ 2 : mean 2 stars parameter µ : mean edges parameter
38 This statement can be quantified in a number of ways: Result: Let e be a unit vector in R q and bd(e) = sup µ intc (e T µ). µ(λe) bd(e)e as λ. 2 P λe,y (Y deg Y) as λ. 3 For every d < bd(e), P λe,y (e T Z (Y ) d) 0 as λ. 4 Let η 0 intc. Then Kullback Leibler divergence(η 0 ; λe) as λ.
39 Effect of Near-Degeneracy on MCMC Estimation Closely related to nice properties of simple MCMC schemes (Geyer ). If a random graph model is simulated using a MCMC based on a near-degenerate ψ it will very likely fail.
40 Effect of Near-Degeneracy on MCMC Estimation Closely related to nice properties of simple MCMC schemes (Geyer ). If a random graph model is simulated using a MCMC based on a near-degenerate ψ it will very likely fail. Full-conditional MCMC with dyad update: M(ψ) = max y Y ψt δ(y c ij ) where δ(yij c) = Z (y + ij ) Z (y ij ) As µ(ψ) bd(c), M(ψ) There exists y Y with [ ] logit P(Y ij = Yij c = yij c ) = ±M(ψ) If ψ is near-degenerate then M(ψ) is large and the MCMC will mix very slowly.
41 Example of degeneracy of the 2 star model P(Y = y) = exp{η E(y) + η 2 S(y)} c(η, η 2 ) M(η) = max{ η, η + 2(g 2)η 2 } y Y
42 Example of degeneracy of the 2 star model P(Y = y) = exp{η E(y) + η 2 S(y)} c(η, η 2 ) y Y M(η) = max{ η, η + 2(g 2)η 2 } MCMC will usually mix poorly.
43 Example of degeneracy of the 2 star model P(Y = y) = exp{η E(y) + η 2 S(y)} c(η, η 2 ) y Y M(η) = max{ η, η + 2(g 2)η 2 } MCMC will usually mix poorly. If µ(η) close to (3, 0) (e.g., η = (4.5,.4)) then M(η) = 4.5 So an MCMC will approach (3, 0) and stay there (.% and.% at (2, 0) bd(c)).
44 Example of degeneracy of the 2 star model P(Y = y) = exp{η E(y) + η 2 S(y)} c(η, η 2 ) y Y M(η) = max{ η, η + 2(g 2)η 2 } MCMC will usually mix poorly. If µ(η) close to (3, 0) (e.g., η = (4.5,.4)) then M(η) = 4.5 So an MCMC will approach (3, 0) and stay there (.% and.% at (2, 0) bd(c)). If µ(η) close to (, 40) (e.g., η = ( 3.43, 0.63)) then M(η) = The model places 50% of its mass on graphs with 2 or fewer edges and 36% on graphs with at least edges.
45 Example of degeneracy of the 2 star model P(Y = y) = exp{η E(y) + η 2 S(y)} c(η, η 2 ) y Y M(η) = max{ η, η + 2(g 2)η 2 } MCMC will usually mix poorly. If µ(η) close to (3, 0) (e.g., η = (4.5,.4)) then M(η) = 4.5 So an MCMC will approach (3, 0) and stay there (.% and.% at (2, 0) bd(c)). If µ(η) close to (, 40) (e.g., η = ( 3.43, 0.63)) then M(η) = The model places 50% of its mass on graphs with 2 or fewer edges and 36% on graphs with at least edges. The model is also unstable e.g., η = ( 3.43, 0.6)) µ(η) (4.4,.) and the model places almost all its mass on empty graphs.
46 Figure 4: Regions of the parameter space of µ µ 2 : mean 2 stars parameter µ : mean edges parameter
47 (a) Trace plot of edges (b) Density of edges Density e+04 6e+04 e+05 Iterations edges (c) Trace plot of 2 stars (d) Density of 2 stars Density e+04 6e+04 e+05 Iterations stars
48 Estimation within the mean value parametization If Z (y obs ) int(c), the MLE of µ is Z (y obs ).
49 Estimation within the mean value parametization If Z (y obs ) int(c), the MLE of µ is Z (y obs ). If Z (y obs ) int(c) the MLE of µ does not exist.
50 Estimation within the mean value parametization If Z (y obs ) int(c), the MLE of µ is Z (y obs ). If Z (y obs ) int(c) the MLE of µ does not exist. The MLE ˆµ is unbiased and has minimum variance: [ ] log c(η) E η (ˆµ) = E η [Z (Y )] = µ(η) = (η) η i [ 2 ] log c(η) V η (ˆµ) = V η [Z (Y )] = (η) η i η j An estimate of the variance-covariance is available using the same MCMC.
51 Trace plot of edges Density of edges Density e+04 6e+04 e+05 Iterations edges Trace plot of 2 stars Density of 2 stars Density e+04 6e+04 e+05 Iterations stars
52 Figure 4: Regions of the parameter space of µ µ 2 : mean 2 stars parameter µ : mean edges parameter
53 Existence and uniqueness of MLE Let C be the convex hull of {Z (y) : y Y} - the convex hull of the discrete support points. Let int(c) be the interior of C.
54 Existence and uniqueness of MLE Let C be the convex hull of {Z (y) : y Y} - the convex hull of the discrete support points. Let int(c) be the interior of C. Result (Barndorff-Nielsen ) The MLE exists if, and only if, Z (y observed ) int(c) If it exists, it is unique and can be found by solving the likelihood equations or by direct optimization of L.
55 Figure : Enumeration of sufficient statistics for graphs with nodes. The circles are centered on the possible values and the area of the circle is proportional to the number of graphs with that value of the sufficient statistic. There are a total of 2,0,52 graphs. number of 2 stars number of edges
56 How can we tell if a model class is useful? Many aspects: Is the model-class itself able to represent a range of realistic networks? model degeneracy: small range of graphs covered as the parameters vary (Handcock 2003) What are the properties of different methods of estimation? e.g, MLE, psuedolikelihood, Bayesian framework computational failure: estimates do not exist for certain observable graphs Can we assess the goodness-of-fit of models? appropriate measures and tests (Besag 2000; Hunter, Goodreau, Handcock 200)
57 Existence and uniqueness of MC-MLE Geyer and Thompson (2) show the MC-MLE converges to the true MLE as the number of simulations increases. also produces estimates of the asymptotic covariance matrix, size of the MCMC induced error, etc. Let CO be the convex hull of sampled sufficient statistics. In practice, three cases: Z (y) int(co) C: MC-MLE exists and is unique 2 Z (y) int(co) but is in int(c): MC-MLE does not exist, even though MLE does 3 Z (y) int(c): MC-MLE and MLE do not exist
58 Figure : Enumeration of sufficient statistics for graphs with nodes. The circles are centered on the possible values and the area of the circle is proportional to the number of graphs with that value of the sufficient statistic. There are a total of 2,0,52 graphs. number of 2 stars number of edges
59 How can we tell if a model class is useful? Many aspects: Is the model-class itself able to represent a range of realistic networks? model degeneracy: small range of graphs covered as the parameters vary (Handcock 2003) What are the properties of different methods of estimation? e.g, MLE, psuedolikelihood, Bayesian framework computational failure: estimates do not exist for certain observable graphs Can we assess the goodness-of-fit of models? appropriate measures and tests (Besag 2000; Hunter, Goodreau, Handcock 200)
60 Goodness of fit intuition ERGM class exp{η g(y)}
61 Goodness of fit intuition ERGM class exp{η g(y)} y obs
62 Goodness of fit intuition ERGM (approx) class MLE exp{η g(y)} η y obs
63 Goodness of fit intuition ERGM (approx) Fitted class MLE ERGM exp{η g(y)} η exp{ η g(y)} y obs
64 Goodness of fit intuition ERGM (approx) Fitted class MLE ERGM exp{η g(y)} η exp{ η g(y)} y obs Randomly generated networks Ỹ, Ỹ2,...
65 Goodness of fit intuition ERGM (approx) Fitted class MLE ERGM exp{η g(y)} η exp{ η g(y)} y obs Randomly generated networks Ỹ, Ỹ2,... Question: How does y obs look as a representative of the sample Ỹ, Ỹ2,...?
66 The eyeball test The data: School 0: 205 Students Simulated network, model A: Simulated graph: By grade 0 2??
67 The eyeball test (cont d) The data: School 0: 205 Students Simulated network, model B: Simulated graph: By grade 0 2?? (Yikes!)
68 The models Model A: g(y) contains terms for # of edges Homophily effects of grade, sex, and race factors Main effects of grade, sex, and race factors i (.632)i EP i, where EP i =# edges with i shared partners Model B: g(y) contains terms for # of edges # of neighbors of the same sex (homophily effect) # of 2-stars # of triangles (Note: It was necessary to use MPLE to fit Model B)
69 Quantitative checks for goodness of fit A well-known example: Lamberteschi Pucci Guadagni Tornabuoni Albizzi Ginori Medici Acciaiuoli Salviati Bischeri Peruzzi Strozzi Castellani Ridolfi Barbadori Florentine marriage data Edge indicates marriage tie between families Sides=degree + 3 Color=degree Size=log(wealth) Pazzi
70 Quantitative checks for goodness of fit A well-known example: Lamberteschi Pucci Guadagni Tornabuoni Albizzi Ginori Medici Acciaiuoli Salviati Bischeri Peruzzi Strozzi Castellani Ridolfi Barbadori Florentine marriage data Edge indicates marriage tie between families Sides=degree + 3 Color=degree Size=log(wealth) Pazzi model <- ergm(flomarriage ~ edges + kstar(2))
71 Graphical GOF check: degree distribution model <- ergm(flomarriage ~ edges + kstar(2)) Goodness of fit diagnostics proportion of nodes degree
72 Graphical GOF: edgewise shared partner distribution model <- ergm(flomarriage ~ edges + kstar(2)) Goodness of fit diagnostics proportion of edges edge wise shared partners
73 Graphical GOF check: geodesic distance distribution model <- ergm(flomarriage ~ edges + kstar(2)) Goodness of fit diagnostics proportion of dyads NR minimum geodesic distance
74 GOF check: Examples from Add Health networks n=205 log odds for a node log odds for an edge log odds for a dyad log odds for a dyad degree log odds for a dyad edge wise shared partners n=220 School 44, Edges + Attributes + WSPartner(.5), free degree log odds for a dyad minimum geodesic distance degree shared partners minimum geodesic distance
75 MCMC Testing Significance tests based on comparing the observed value of a statistics to a null probability distribution. MCMC p values (2000) Besag and Clifford (), Besag
76 Illustration: Empirical evidence of competition among Darwin s Finches Island Finch A B C D E F G H I J K L M N O P Q Large ground finch Medium ground finch Small ground finch Sharp-beaked ground finch Cactus ground finch Large cactus ground finch Large tree finch Medium tree finch Small tree finch Vegetarian finch Woodpecker finch Mangrove finch Warbler finch Table: Darwin s finch data
77 MCMC Testing and p values Does the observed grouping of finch species on islands happened by random chance or if it was the result of a struggle in which only species which depended on different food sources could coexist on an island. To test this hypothesis, consider the test statistic S 2 = m(m ) sij 2, where m is the number of finch species, S = ( s ij ) = AA T, and A = ( a ij ) is the bipartite graph in the table. i j
78 Figure: Null distribution of the test statistic S 2
79 Figure: Number of pairs of finches sharing x islands, x = 0,,...,
80 Summary Network representations intersect with most sciences Sparse models are being used to capture structural properties The models must depend on the scientific objective. Some seemingly simple models are not so. The inclusion of attributes is very important actor attributes dyad attributes e.g. homophily, race, location structural terms e.g. transitive homophily
81 We need better and more local models for social networks: e.g. nearest neighbor" ideas for local dependence Baddeley and Moller () Snijders, Robins, Pattison, Handcock (2006) Taking into account class membership is very important known classes block models" Wang and Wong ()
82 latent class and trait models are important an underlying latent social space" of actors Hoff, Raftery and Handcock (2002) Hoff (2003, 2004,...) latent class models are very promising Nowicki and Snijders (200) latent class and trait models Handcock, Raftery, Tantrum (200); Krivitsky et. al (200) Hoff (2005, 200) grade of membership models Airoldi, Blei, Feinberg (200)
Statistical Models for Social Networks with Application to HIV Epidemiology
Statistical Models for Social Networks with Application to HIV Epidemiology Mark S. Handcock Department of Statistics University of Washington Joint work with Pavel Krivitsky Martina Morris and the U.
More informationCurved exponential family models for networks
Curved exponential family models for networks David R. Hunter, Penn State University Mark S. Handcock, University of Washington February 18, 2005 Available online as Penn State Dept. of Statistics Technical
More informationAssessing Goodness of Fit of Exponential Random Graph Models
International Journal of Statistics and Probability; Vol. 2, No. 4; 2013 ISSN 1927-7032 E-ISSN 1927-7040 Published by Canadian Center of Science and Education Assessing Goodness of Fit of Exponential Random
More informationAlgorithmic approaches to fitting ERG models
Ruth Hummel, Penn State University Mark Handcock, University of Washington David Hunter, Penn State University Research funded by Office of Naval Research Award No. N00014-08-1-1015 MURI meeting, April
More informationc Copyright 2015 Ke Li
c Copyright 2015 Ke Li Degeneracy, Duration, and Co-evolution: Extending Exponential Random Graph Models (ERGM) for Social Network Analysis Ke Li A dissertation submitted in partial fulfillment of the
More informationSequential Importance Sampling for Bipartite Graphs With Applications to Likelihood-Based Inference 1
Sequential Importance Sampling for Bipartite Graphs With Applications to Likelihood-Based Inference 1 Ryan Admiraal University of Washington, Seattle Mark S. Handcock University of Washington, Seattle
More informationGoodness of Fit of Social Network Models 1
Goodness of Fit of Social Network Models David R. Hunter Pennsylvania State University, University Park Steven M. Goodreau University of Washington, Seattle Mark S. Handcock University of Washington, Seattle
More informationGoodness of Fit of Social Network Models
Goodness of Fit of Social Network Models David R. HUNTER, StevenM.GOODREAU, and Mark S. HANDCOCK We present a systematic examination of a real network data set using maximum likelihood estimation for exponential
More informationInstability, Sensitivity, and Degeneracy of Discrete Exponential Families
Instability, Sensitivity, and Degeneracy of Discrete Exponential Families Michael Schweinberger Abstract In applications to dependent data, first and foremost relational data, a number of discrete exponential
More informationCentrality II. MGT 780 Social Network Analysis Steve Borgatti LINKS Center University of Kentucky. 11 April 2016 (c) 2016 Stephen P Borgatti 1
Centrality II MGT 780 Social Network Analysis Steve Borgatti LINKS Center University of Kentucky 1 Experimental exchange theory Recruit subjects to play game in which pairs of people have to allocate 24
More informationarxiv: v1 [stat.me] 3 Apr 2017
A two-stage working model strategy for network analysis under Hierarchical Exponential Random Graph Models Ming Cao University of Texas Health Science Center at Houston ming.cao@uth.tmc.edu arxiv:1704.00391v1
More informationInference in curved exponential family models for networks
Inference in curved exponential family models for networks David R. Hunter, Penn State University Mark S. Handcock, University of Washington Penn State Department of Statistics Technical Report No. TR0402
More informationarxiv: v1 [stat.me] 1 Aug 2012
Exponential-family Random Network Models arxiv:208.02v [stat.me] Aug 202 Ian Fellows and Mark S. Handcock University of California, Los Angeles, CA, USA Summary. Random graphs, where the connections between
More informationRandom Effects Models for Network Data
Random Effects Models for Network Data Peter D. Hoff 1 Working Paper no. 28 Center for Statistics and the Social Sciences University of Washington Seattle, WA 98195-4320 January 14, 2003 1 Department of
More informationSpecification and estimation of exponential random graph models for social (and other) networks
Specification and estimation of exponential random graph models for social (and other) networks Tom A.B. Snijders University of Oxford March 23, 2009 c Tom A.B. Snijders (University of Oxford) Models for
More informationSampling and incomplete network data
1/58 Sampling and incomplete network data 567 Statistical analysis of social networks Peter Hoff Statistics, University of Washington 2/58 Network sampling methods It is sometimes difficult to obtain a
More informationConditional Marginalization for Exponential Random Graph Models
Conditional Marginalization for Exponential Random Graph Models Tom A.B. Snijders January 21, 2010 To be published, Journal of Mathematical Sociology University of Oxford and University of Groningen; this
More informationStatistical Model for Soical Network
Statistical Model for Soical Network Tom A.B. Snijders University of Washington May 29, 2014 Outline 1 Cross-sectional network 2 Dynamic s Outline Cross-sectional network 1 Cross-sectional network 2 Dynamic
More informationBayesian Analysis of Network Data. Model Selection and Evaluation of the Exponential Random Graph Model. Dissertation
Bayesian Analysis of Network Data Model Selection and Evaluation of the Exponential Random Graph Model Dissertation Presented to the Faculty for Social Sciences, Economics, and Business Administration
More informationModeling tie duration in ERGM-based dynamic network models
University of Wollongong Research Online Faculty of Engineering and Information Sciences - Papers: Part A Faculty of Engineering and Information Sciences 2012 Modeling tie duration in ERGM-based dynamic
More informationFast Maximum Likelihood estimation via Equilibrium Expectation for Large Network Data
Fast Maximum Likelihood estimation via Equilibrium Expectation for Large Network Data Maksym Byshkin 1, Alex Stivala 4,1, Antonietta Mira 1,3, Garry Robins 2, Alessandro Lomi 1,2 1 Università della Svizzera
More informationReview and continuation from last week Properties of MLEs
Review and continuation from last week Properties of MLEs As we have mentioned, MLEs have a nice intuitive property, and as we have seen, they have a certain equivariance property. We will see later that
More informationEstimation for Dyadic-Dependent Exponential Random Graph Models
JOURNAL OF ALGEBRAIC STATISTICS Vol. 5, No. 1, 2014, 39-63 ISSN 1309-3452 www.jalgstat.com Estimation for Dyadic-Dependent Exponential Random Graph Models Xiaolin Yang, Alessandro Rinaldo, Stephen E. Fienberg
More informationGeneralized Exponential Random Graph Models: Inference for Weighted Graphs
Generalized Exponential Random Graph Models: Inference for Weighted Graphs James D. Wilson University of North Carolina at Chapel Hill June 18th, 2015 Political Networks, 2015 James D. Wilson GERGMs for
More informationParameterizing Exponential Family Models for Random Graphs: Current Methods and New Directions
Carter T. Butts p. 1/2 Parameterizing Exponential Family Models for Random Graphs: Current Methods and New Directions Carter T. Butts Department of Sociology and Institute for Mathematical Behavioral Sciences
More informationUnit 5: Centrality. ICPSR University of Michigan, Ann Arbor Summer 2015 Instructor: Ann McCranie
Unit 5: Centrality ICPSR University of Michigan, Ann Arbor Summer 2015 Instructor: Ann McCranie What does centrality tell us? We often want to know who the most important actors in a network are. Centrality
More informationA Graphon-based Framework for Modeling Large Networks
A Graphon-based Framework for Modeling Large Networks Ran He Submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy in the Graduate School of Arts and Sciences COLUMBIA
More informationTEMPORAL EXPONENTIAL- FAMILY RANDOM GRAPH MODELING (TERGMS) WITH STATNET
1 TEMPORAL EXPONENTIAL- FAMILY RANDOM GRAPH MODELING (TERGMS) WITH STATNET Prof. Steven Goodreau Prof. Martina Morris Prof. Michal Bojanowski Prof. Mark S. Handcock Source for all things STERGM Pavel N.
More informationIntroduction to statistical analysis of Social Networks
The Social Statistics Discipline Area, School of Social Sciences Introduction to statistical analysis of Social Networks Mitchell Centre for Network Analysis Johan Koskinen http://www.ccsr.ac.uk/staff/jk.htm!
More informationModel-Based Clustering for Social Networks
Model-Based Clustering for Social Networks Mark S. Handcock, Adrian E. Raftery and Jeremy M. Tantrum University of Washington Technical Report no. 42 Department of Statistics University of Washington April
More informationChaos, Complexity, and Inference (36-462)
Chaos, Complexity, and Inference (36-462) Lecture 21 Cosma Shalizi 3 April 2008 Models of Networks, with Origin Myths Erdős-Rényi Encore Erdős-Rényi with Node Types Watts-Strogatz Small World Graphs Exponential-Family
More informationBayesian Inference for Contact Networks Given Epidemic Data
Bayesian Inference for Contact Networks Given Epidemic Data Chris Groendyke, David Welch, Shweta Bansal, David Hunter Departments of Statistics and Biology Pennsylvania State University SAMSI, April 17,
More informationExponential random graph models for the Japanese bipartite network of banks and firms
Exponential random graph models for the Japanese bipartite network of banks and firms Abhijit Chakraborty, Hazem Krichene, Hiroyasu Inoue, and Yoshi Fujiwara Graduate School of Simulation Studies, The
More informationModeling homophily and stochastic equivalence in symmetric relational data
Modeling homophily and stochastic equivalence in symmetric relational data Peter D. Hoff Departments of Statistics and Biostatistics University of Washington Seattle, WA 98195-4322. hoff@stat.washington.edu
More informationChaos, Complexity, and Inference (36-462)
Chaos, Complexity, and Inference (36-462) Lecture 21: More Networks: Models and Origin Myths Cosma Shalizi 31 March 2009 New Assignment: Implement Butterfly Mode in R Real Agenda: Models of Networks, with
More informationAn approximation method for improving dynamic network model fitting
University of Wollongong Research Online Faculty of Engineering and Information Sciences - Papers: Part A Faculty of Engineering and Information Sciences 2014 An approximation method for improving dynamic
More informationStatistical Methods for Handling Incomplete Data Chapter 2: Likelihood-based approach
Statistical Methods for Handling Incomplete Data Chapter 2: Likelihood-based approach Jae-Kwang Kim Department of Statistics, Iowa State University Outline 1 Introduction 2 Observed likelihood 3 Mean Score
More informationTied Kronecker Product Graph Models to Capture Variance in Network Populations
Tied Kronecker Product Graph Models to Capture Variance in Network Populations Sebastian Moreno, Sergey Kirshner +, Jennifer Neville +, SVN Vishwanathan + Department of Computer Science, + Department of
More informationarxiv: v1 [stat.me] 5 Mar 2013
Analysis of Partially Observed Networks via Exponentialfamily Random Network Models arxiv:303.29v [stat.me] 5 Mar 203 Ian E. Fellows Mark S. Handcock Department of Statistics, University of California,
More information10708 Graphical Models: Homework 2
10708 Graphical Models: Homework 2 Due Monday, March 18, beginning of class Feburary 27, 2013 Instructions: There are five questions (one for extra credit) on this assignment. There is a problem involves
More informationModel comparison and selection
BS2 Statistical Inference, Lectures 9 and 10, Hilary Term 2008 March 2, 2008 Hypothesis testing Consider two alternative models M 1 = {f (x; θ), θ Θ 1 } and M 2 = {f (x; θ), θ Θ 2 } for a sample (X = x)
More informationMarkov Random Fields
Markov Random Fields 1. Markov property The Markov property of a stochastic sequence {X n } n 0 implies that for all n 1, X n is independent of (X k : k / {n 1, n, n + 1}), given (X n 1, X n+1 ). Another
More informationHierarchical Models for Social Networks
Hierarchical Models for Social Networks Tracy M. Sweet University of Maryland Innovative Assessment Collaboration November 4, 2014 Acknowledgements Program for Interdisciplinary Education Research (PIER)
More informationOverview course module Stochastic Modelling
Overview course module Stochastic Modelling I. Introduction II. Actor-based models for network evolution III. Co-evolution models for networks and behaviour IV. Exponential Random Graph Models A. Definition
More informationMaster s Written Examination
Master s Written Examination Option: Statistics and Probability Spring 05 Full points may be obtained for correct answers to eight questions Each numbered question (which may have several parts) is worth
More informationDelayed Rejection Algorithm to Estimate Bayesian Social Networks
Dublin Institute of Technology ARROW@DIT Articles School of Mathematics 2014 Delayed Rejection Algorithm to Estimate Bayesian Social Networks Alberto Caimo Dublin Institute of Technology, alberto.caimo@dit.ie
More informationUndirected Graphical Models
Undirected Graphical Models 1 Conditional Independence Graphs Let G = (V, E) be an undirected graph with vertex set V and edge set E, and let A, B, and C be subsets of vertices. We say that C separates
More informationIV. Analyse de réseaux biologiques
IV. Analyse de réseaux biologiques Catherine Matias CNRS - Laboratoire de Probabilités et Modèles Aléatoires, Paris catherine.matias@math.cnrs.fr http://cmatias.perso.math.cnrs.fr/ ENSAE - 2014/2015 Sommaire
More informationAn Approximation Method for Improving Dynamic Network Model Fitting
DEPARTMENT OF STATISTICS The Pennsylvania State University University Park, PA 16802 U.S.A. TECHNICAL REPORTS AND PREPRINTS Number 12-04: October 2012 An Approximation Method for Improving Dynamic Network
More informationAn Introduction to Exponential-Family Random Graph Models
An Introduction to Exponential-Family Random Graph Models Luo Lu Feb.8, 2011 1 / 11 Types of complications in social network Single relationship data A single relationship observed on a set of nodes at
More informationAsymptotics and Extremal Properties of the Edge-Triangle Exponential Random Graph Model
Asymptotics and Extremal Properties of the Edge-Triangle Exponential Random Graph Model Alessandro Rinaldo Carnegie Mellon University joint work with Mei Yin, Sukhada Fadnavis, Stephen E. Fienberg and
More informationBargaining, Information Networks and Interstate
Bargaining, Information Networks and Interstate Conflict Erik Gartzke Oliver Westerwinter UC, San Diego Department of Political Sciene egartzke@ucsd.edu European University Institute Department of Political
More informationNetwork Event Data over Time: Prediction and Latent Variable Modeling
Network Event Data over Time: Prediction and Latent Variable Modeling Padhraic Smyth University of California, Irvine Machine Learning with Graphs Workshop, July 25 th 2010 Acknowledgements PhD students:
More informationModeling heterogeneity in random graphs
Modeling heterogeneity in random graphs Catherine MATIAS CNRS, Laboratoire Statistique & Génome, Évry (Soon: Laboratoire de Probabilités et Modèles Aléatoires, Paris) http://stat.genopole.cnrs.fr/ cmatias
More informationExponential Families
Exponential Families David M. Blei 1 Introduction We discuss the exponential family, a very flexible family of distributions. Most distributions that you have heard of are in the exponential family. Bernoulli,
More informationDiscrete Temporal Models of Social Networks
Discrete Temporal Models of Social Networks Steve Hanneke and Eric Xing Machine Learning Department Carnegie Mellon University Pittsburgh, PA 15213 USA {shanneke,epxing}@cs.cmu.edu Abstract. We propose
More informationModeling of Dynamic Networks based on Egocentric Data with Durational Information
DEPARTMENT OF STATISTICS The Pennsylvania State University University Park, PA 16802 U.S.A. TECHNICAL REPORTS AND PREPRINTS Number 12-01: April 2012 Modeling of Dynamic Networks based on Egocentric Data
More informationProbabilistic Graphical Models
2016 Robert Nowak Probabilistic Graphical Models 1 Introduction We have focused mainly on linear models for signals, in particular the subspace model x = Uθ, where U is a n k matrix and θ R k is a vector
More informationMaintaining Nets and Net Trees under Incremental Motion
Maintaining Nets and Net Trees under Incremental Motion Minkyoung Cho, David Mount, and Eunhui Park Department of Computer Science University of Maryland, College Park August 25, 2009 Latent Space Embedding
More informationEfficient MCMC Samplers for Network Tomography
Efficient MCMC Samplers for Network Tomography Martin Hazelton 1 Institute of Fundamental Sciences Massey University 7 December 2015 1 Email: m.hazelton@massey.ac.nz AUT Mathematical Sciences Symposium
More informationUsing Potential Games to Parameterize ERG Models
Carter T. Butts p. 1/2 Using Potential Games to Parameterize ERG Models Carter T. Butts Department of Sociology and Institute for Mathematical Behavioral Sciences University of California, Irvine buttsc@uci.edu
More informationStat 5101 Lecture Notes
Stat 5101 Lecture Notes Charles J. Geyer Copyright 1998, 1999, 2000, 2001 by Charles J. Geyer May 7, 2001 ii Stat 5101 (Geyer) Course Notes Contents 1 Random Variables and Change of Variables 1 1.1 Random
More informationEXAM IN STATISTICAL MACHINE LEARNING STATISTISK MASKININLÄRNING
EXAM IN STATISTICAL MACHINE LEARNING STATISTISK MASKININLÄRNING DATE AND TIME: August 30, 2018, 14.00 19.00 RESPONSIBLE TEACHER: Niklas Wahlström NUMBER OF PROBLEMS: 5 AIDING MATERIAL: Calculator, mathematical
More informationA graph contains a set of nodes (vertices) connected by links (edges or arcs)
BOLTZMANN MACHINES Generative Models Graphical Models A graph contains a set of nodes (vertices) connected by links (edges or arcs) In a probabilistic graphical model, each node represents a random variable,
More informationLoglikelihood and Confidence Intervals
Stat 504, Lecture 2 1 Loglikelihood and Confidence Intervals The loglikelihood function is defined to be the natural logarithm of the likelihood function, l(θ ; x) = log L(θ ; x). For a variety of reasons,
More informationBayesian Methods for Machine Learning
Bayesian Methods for Machine Learning CS 584: Big Data Analytics Material adapted from Radford Neal s tutorial (http://ftp.cs.utoronto.ca/pub/radford/bayes-tut.pdf), Zoubin Ghahramni (http://hunch.net/~coms-4771/zoubin_ghahramani_bayesian_learning.pdf),
More informationManaging Uncertainty
Managing Uncertainty Bayesian Linear Regression and Kalman Filter December 4, 2017 Objectives The goal of this lab is multiple: 1. First it is a reminder of some central elementary notions of Bayesian
More informationCSC 412 (Lecture 4): Undirected Graphical Models
CSC 412 (Lecture 4): Undirected Graphical Models Raquel Urtasun University of Toronto Feb 2, 2016 R Urtasun (UofT) CSC 412 Feb 2, 2016 1 / 37 Today Undirected Graphical Models: Semantics of the graph:
More informationConfidence Sets for Network Structure
Confidence Sets for Network Structure Edoardo M. Airoldi 1, David S. Choi 2 and Patrick J. Wolfe 1 1 Department of Statistics, Harvard University 2 School of Engineering and Applied Sciences, Harvard University
More informationDepartment of Statistics. Bayesian Modeling for a Generalized Social Relations Model. Tyler McCormick. Introduction.
A University of Connecticut and Columbia University A models for dyadic data are extensions of the (). y i,j = a i + b j + γ i,j (1) Here, y i,j is a measure of the tie from actor i to actor j. The random
More informationNetworks in the Real World
Networks: Lecture 2 Introduction Networks in the Real World A network is a set of items (nodes or vertices) connected by edges or links. Systems taking the form of networks abound in the world. Types of
More informationA multilayer exponential random graph modelling approach for weighted networks
A multilayer exponential random graph modelling approach for weighted networks Alberto Caimo 1 and Isabella Gollini 2 1 Dublin Institute of Technology, Ireland; alberto.caimo@dit.ie arxiv:1811.07025v1
More informationi=1 h n (ˆθ n ) = 0. (2)
Stat 8112 Lecture Notes Unbiased Estimating Equations Charles J. Geyer April 29, 2012 1 Introduction In this handout we generalize the notion of maximum likelihood estimation to solution of unbiased estimating
More informationBayesian Melding. Assessing Uncertainty in UrbanSim. University of Washington
Bayesian Melding Assessing Uncertainty in UrbanSim Hana Ševčíková University of Washington hana@stat.washington.edu Joint work with Paul Waddell and Adrian Raftery University of Washington UrbanSim Workshop,
More informationarxiv: v1 [stat.me] 12 Apr 2018
A New Generative Statistical Model for Graphs: The Latent Order Logistic (LOLOG) Model Ian E. Fellows Fellows Statistics, San Diego, United States E-mail: ian@fellstat.com arxiv:1804.04583v1 [stat.me]
More informationThe Origin of Deep Learning. Lili Mou Jan, 2015
The Origin of Deep Learning Lili Mou Jan, 2015 Acknowledgment Most of the materials come from G. E. Hinton s online course. Outline Introduction Preliminary Boltzmann Machines and RBMs Deep Belief Nets
More informationDynamic modeling of organizational coordination over the course of the Katrina disaster
Dynamic modeling of organizational coordination over the course of the Katrina disaster Zack Almquist 1 Ryan Acton 1, Carter Butts 1 2 Presented at MURI Project All Hands Meeting, UCI April 24, 2009 1
More informationLearning with Blocks: Composite Likelihood and Contrastive Divergence
Arthur U. Asuncion 1, Qiang Liu 1, Alexander T. Ihler, Padhraic Smyth {asuncion,qliu1,ihler,smyth}@ics.uci.edu Department of Computer Science University of California, Irvine Abstract Composite likelihood
More informationContinuous-time Statistical Models for Network Panel Data
Continuous-time Statistical Models for Network Panel Data Tom A.B. Snijders University of Groningen University of Oxford September, 2016 1 / 45 Overview 1 Models for network panel data 2 Example 3 Co-evolution
More informationMassive-scale estimation of exponential-family random graph models with local dependence
Massive-scale estimation of exponential-family random graph models with local dependence Sergii Babkin Michael Schweinberger arxiv:1703.09301v1 [stat.co] 27 Mar 2017 Abstract A flexible approach to modeling
More informationChapter 17: Undirected Graphical Models
Chapter 17: Undirected Graphical Models The Elements of Statistical Learning Biaobin Jiang Department of Biological Sciences Purdue University bjiang@purdue.edu October 30, 2014 Biaobin Jiang (Purdue)
More informationBrandon C. Kelly (Harvard Smithsonian Center for Astrophysics)
Brandon C. Kelly (Harvard Smithsonian Center for Astrophysics) Probability quantifies randomness and uncertainty How do I estimate the normalization and logarithmic slope of a X ray continuum, assuming
More informationGaussian Processes 1. Schedule
1 Schedule 17 Jan: Gaussian processes (Jo Eidsvik) 24 Jan: Hands-on project on Gaussian processes (Team effort, work in groups) 31 Jan: Latent Gaussian models and INLA (Jo Eidsvik) 7 Feb: Hands-on project
More informationLikelihood Inference for Lattice Spatial Processes
Likelihood Inference for Lattice Spatial Processes Donghoh Kim November 30, 2004 Donghoh Kim 1/24 Go to 1234567891011121314151617 FULL Lattice Processes Model : The Ising Model (1925), The Potts Model
More informationAppendix: Modeling Approach
AFFECTIVE PRIMACY IN INTRAORGANIZATIONAL TASK NETWORKS Appendix: Modeling Approach There is now a significant and developing literature on Bayesian methods in social network analysis. See, for instance,
More informationMultilevel Statistical Models: 3 rd edition, 2003 Contents
Multilevel Statistical Models: 3 rd edition, 2003 Contents Preface Acknowledgements Notation Two and three level models. A general classification notation and diagram Glossary Chapter 1 An introduction
More informationFractional Imputation in Survey Sampling: A Comparative Review
Fractional Imputation in Survey Sampling: A Comparative Review Shu Yang Jae-Kwang Kim Iowa State University Joint Statistical Meetings, August 2015 Outline Introduction Fractional imputation Features Numerical
More informationHigh dimensional Ising model selection
High dimensional Ising model selection Pradeep Ravikumar UT Austin (based on work with John Lafferty, Martin Wainwright) Sparse Ising model US Senate 109th Congress Banerjee et al, 2008 Estimate a sparse
More informationUndirected Graphical Models
Outline Hong Chang Institute of Computing Technology, Chinese Academy of Sciences Machine Learning Methods (Fall 2012) Outline Outline I 1 Introduction 2 Properties Properties 3 Generative vs. Conditional
More informationStatistics 360/601 Modern Bayesian Theory
Statistics 360/601 Modern Bayesian Theory Alexander Volfovsky Lecture 8 - Sept 25, 2018 Monte Carlo 1 2 Monte Carlo approximation Want to compute Z E[ y] = p( y)d without worrying about actual integration...
More informationStat 542: Item Response Theory Modeling Using The Extended Rank Likelihood
Stat 542: Item Response Theory Modeling Using The Extended Rank Likelihood Jonathan Gruhl March 18, 2010 1 Introduction Researchers commonly apply item response theory (IRT) models to binary and ordinal
More informationConsistency Under Sampling of Exponential Random Graph Models
Consistency Under Sampling of Exponential Random Graph Models Cosma Shalizi and Alessandro Rinaldo Summary by: Elly Kaizar Remember ERGMs (Exponential Random Graph Models) Exponential family models Sufficient
More informationProperties of Latent Variable Network Models
Properties of Latent Variable Network Models Riccardo Rastelli and Nial Friel University College Dublin Adrian E. Raftery University of Washington Technical Report no. 634 Department of Statistics University
More informationGraphical Models and Kernel Methods
Graphical Models and Kernel Methods Jerry Zhu Department of Computer Sciences University of Wisconsin Madison, USA MLSS June 17, 2014 1 / 123 Outline Graphical Models Probabilistic Inference Directed vs.
More informationChris Bishop s PRML Ch. 8: Graphical Models
Chris Bishop s PRML Ch. 8: Graphical Models January 24, 2008 Introduction Visualize the structure of a probabilistic model Design and motivate new models Insights into the model s properties, in particular
More informationEstimating the Size of Hidden Populations using Respondent-Driven Sampling Data
Estimating the Size of Hidden Populations using Respondent-Driven Sampling Data Mark S. Handcock Krista J. Gile Department of Statistics Department of Mathematics University of California University of
More informationGeneralized Linear Models
Generalized Linear Models Lecture 3. Hypothesis testing. Goodness of Fit. Model diagnostics GLM (Spring, 2018) Lecture 3 1 / 34 Models Let M(X r ) be a model with design matrix X r (with r columns) r n
More informationStatistical Data Mining and Machine Learning Hilary Term 2016
Statistical Data Mining and Machine Learning Hilary Term 2016 Dino Sejdinovic Department of Statistics Oxford Slides and other materials available at: http://www.stats.ox.ac.uk/~sejdinov/sdmml Naïve Bayes
More informationProbabilistic Graphical Models & Applications
Probabilistic Graphical Models & Applications Learning of Graphical Models Bjoern Andres and Bernt Schiele Max Planck Institute for Informatics The slides of today s lecture are authored by and shown with
More informationStochastic blockmodels with a growing number of classes
Biometrika (2012), 99,2,pp. 273 284 doi: 10.1093/biomet/asr053 C 2012 Biometrika Trust Advance Access publication 17 April 2012 Printed in Great Britain Stochastic blockmodels with a growing number of
More information