Phase Transitions in the Information Distortion NIPS 2003 workshop on Information Theory and Learning: The Bottleneck and Distortion Approach December 13, 2003 Albert E. Parker Department of Mathematical Sciences Center for Computational Biology Montana State University Collaborators: Tomas Gedeon, Alex Dimitrov, John Miller, and Zane Aldworth The Goal: To determine the phase transitions or the bifurcation structure of solutions to clustering problems of the form maxqG(q) constrained by D(q)I0 where q(Z|X) X Z • is the set of valid conditional probabilities in RNK. K objects N clusters • G and D are sufficiently smooth in . • G and D have symmetry: they are invariant to relabelling of the classes of Z. • The Hessians qG and q D are block diagonal. A similar formulation: Using the method Lagrange multipliers, the goal of determining the bifurcation structure of solutions of the optimization problem can be rephrased as finding the bifurcation structure of stationary points of the problem maxq(G(q)+D(q)) where • [0,). • is the set of valid conditional probabilities in RNK. q(Z|X) X K objects Z N clusters • G and D are sufficiently smooth in . • G and D have symmetry: they are invariant to relabelling of the classes of Z. • The Hessian q(G+ D) is block diagonal, and satisfies a set of regularity conditions at bifurcation: (e.g. the kernel of each block is one dimensional) How: Use the Symmetries By capitalizing on the symmetries of the cost functions, we have described the bifurcation structure of stationary points to problems of the form maxqG(q) constrained by D(q)I0 or maxq(G(q)+D(q)) where • [0,). • is the set of valid conditional probabilities in RNK. • G and D are sufficiently smooth in . • G and D have symmetry: they are invariant to relabelling of the classes of Z. • The Hessian q(G+ D) is block diagonal, and satisfies a set of regularity conditions at bifurcation: (e.g. the kernel is one dimensional) Examples optimizing at a distortion level D(Y,Z) D0 • Rate Distortion Theory (Shannon 1950’s) Minimal Informative Compression min I(X,Z) constrained by D(X,Z) D0 q • Deterministic Annealing (Rose 1990’s) A Clustering Algorithm max H(Z|X) constrained by D(X,Z) D0 C, q Examples optimizing at a distortion level D(Y,Z) D0 • Rate Distortion Theory (Shannon 1950’s) Minimal Informative Compression max -I(X,Z) constrained by D(X,Z) D0 q • Deterministic Annealing (Rose 1998) A Clustering Algorithm max H(Z|X) constrained by D(X,Z) D0 C, q I(X,Z)=H(Z) – H(Z|X) Inputs and Outputs and Clustered Outputs Inputs Y L objects {yi} Outputs p(X,Y) X K objects {xi} Clustered Outputs q(Z|X) Z N objects {zi} Inputs and Outputs and Clustered Outputs Inputs Y L objects {yi} Outputs p(X,Y) X K objects {xi} Clustered Outputs q(Z|X) Z N objects {zi} Two methods which use an information distortion function to cluster • Information Bottleneck Method (Tishby, Pereira, Bialek 1999) min I(X,Z) constrained by DI(X,Z) D0 q max –I(X,Z) + I(Y;Z) q • Information Distortion Method (Dimitrov and Miller 2001) max H(Z|X) constrained by DI(X,Z) D0 q max H(Z|X) + I(Y;Z) q Two methods which use an information distortion function to cluster • Information Bottleneck Method (Tishby, Pereira, Bialek 1999) min I(X,Z) constrained by DI(X,Z) D0 q max –I(X,Z) + I(Y;Z) q The Hessian is always singular … (-I(X,Z) is not strictly concave) The theory which follows does not apply • Information Distortion Method (Dimitrov and Miller 2001) max H(Z|X) constrained by DI(X,Z) D0 q max H(Z|X) + I(Y;Z) q Two methods which use an information distortion function to cluster • Information Bottleneck Method (Tishby, Pereira, Bialek 1999) min I(X,Z) constrained by DI(X,Z) D0 q max –I(X,Z) + I(Y;Z) q The Hessian is always singular … (I(X,Z) is not strictly concave) The theory which follows does not apply • Information Distortion Method (Dimitrov and Miller 2001) max H(Z|X) constrained by DI(X,Z) D0 q H(Z|X) is strictly concave) max H(Z|X) + I(Y;Z) q The theory which follows does apply A basic annealing algorithm to solve maxq(G(q)+D(q)) Let q0 be the maximizer of maxq G(q), and let 0 =0. For k 0, let (qk , k ) be a solution to maxq G(q) + D(q ). Iterate the following steps until K = max for some K. 1. Perform -step: Let k+1 = k + dk where dk>0 2. The initial guess for qk+1 at k+1 is qk+1(0) = qk + for some small perturbation . 3. Optimization: solve maxq (G(q) + k+1 D(q)) to get the maximizer qk+1 , using initial guess qk+1(0) . Application of the annealing method to the Information Distortion problem maxq (H(Z|X) + I(X;Z)) when p(X,Y) is defined by four gaussian blobs Y p(X,Y) L=52 inputs X X K=52 outputs X, Outputs Z K=52 outputs N=4 clustered outputs Z, Clustered Outputs Y, Inputs q(Z|X) X, Outputs Evolution of the optimal clustering: Observed Bifurcations for the Four Blob problem: I(Y,Z) bits We just saw the optimal clusterings q* at some *= max . What do the clusterings look like for < max ?? ?????? Observed Bifurcations for the 4 Blob Problem Conceptual Bifurcation Structure I(Y,Z) bits q* Why are there only 3 bifurcations observed? In general, are there only N-1 bifurcations? What kinds of bifurcations do we expect: pitchfork-like, transcritical, saddle-node, or some other type? How many bifurcating branches are there? What do the bifurcating branches look like? Are they 1st order phase transitions (subcritical) or 2nd order phase transitions (supercritical) ? What is the stability of the bifurcating branches? Is there always a bifurcating branch which contains solutions of the optimization problem? Are there bifurcations after all of the classes have resolved ? Recall the Symmetries: To better understand the bifurcation structure, we capitalize on the symmetries of the function G(q)+D(q) class 1 class 3 q(Z|X) : a clustering X Z K objects {xi} N objects {zi} Recall the Symmetries: To better understand the bifurcation structure, we capitalize on the symmetries of the function G(q)+D(q) class 3 class 1 q(Z|X) : a clustering X Z K objects {xi} N objects {zi} The symmetry group of all permutations on N symbols is SN . A partial subgroup lattice for SN when N=4. S4 S3 S2 S2 S3 S2 S2 S2 S3 S2 S2 1 S2 S2 S3 S2 S2 S2 A partial lattice of the maximal subgroups S2 x S2 of S4 S4 12, 34 13, 24 14, 23 This Group Structure determines the Bifurcation Structure Define a Gradient Flow Goal: To determine the bifurcation structure of stationary points of maxq (G(q) + D(q)) Method: Study the equilibria of the of the flow q q , L (q, , ) : q , G(q) D(q) y q( z | x) 1 yY z • Equilibria of this system (in RNK+K ) are possible solutions of the optimization problem • The Jacobian q,L(q*,*) is symmetric, and so only bifurcations of equilibria can occur. • The first equilibrium is q*(0 = 0) 1/N. Symmetry Breaking Bifurcations q* q 1 is fixed by S N S 4 N q1 N 1 4 S4 S3 S2 S2 S2 S3 S2 S2 S3 S2 S2 1 S2 S2 S3 S2 S2 S2 Symmetry Breaking Bifurcations q is fixed by S N 1 S3 * q* q* q 1 is fixed by S N S 4 N q1 N 1 4 S4 S3 S2 S2 S2 S3 S2 S2 S3 S2 S2 1 S2 S2 S3 S2 S2 S2 Symmetry Breaking Bifurcations q is fixed by S N 1 S3 * q* q * is fixed by S N 2 S 2 q* q* q 1 is fixed by S N S 4 N q1 N 1 4 S4 S3 S2 S2 S2 S3 S2 S2 S3 S2 S2 1 S2 S2 S3 S2 S2 S2 Symmetry Breaking Bifurcations q* S4 S3 S2 S2 S2 S3 S2 S2 S3 S2 S2 1 S2 S2 S3 S2 S2 S2 Symmetry Breaking Bifurcations q* S4 12, 34 13, 24 14, 23 Symmetry Breaking Bifurcations q* is fixed by S2 S2 (12), (34) q* q* S4 12, 34 13, 24 14, 23 Existence Theorems for Bifurcating Branches q* Given a bifurcation at a point fixed by SN , • Equivariant Branching Lemma The Smoller-Wasserman Theorem (Vanderbauwhede and Cicogna 1980-1) (Smoller and Wasserman 1985-6) • • There are N bifurcating branches, each which have symmetry SN-1 . There are N!/(2m!n!) bifurcating branches which have symmetry Sm x Sn if N=m+n. Existence Theorems for Bifurcating Branches q* Given a bifurcation at a point fixed by SN-1 , • Equivariant Branching Lemma The Smoller-Wasserman Theorem (Vanderbauwhede and Cicogna 1980-1) (Smoller and Wasserman 1985-6) • • There are N-1 bifurcating branches, each which have symmetry SN-2 . There are (N-1)!/(2m!n!) bifurcating branches which have symmetry Sm x Sn if N-1=m+n. Observed Bifurcation Structure Group Structure S4 S3 S2 S2 S2 S3 S2 S2 S3 S2 S2 1 S2 S2 S3 S2 S2 S2 The Equivariant Branching Lemma shows that the bifurcation structure contains the branches … Observed Bifurcation Structure q* Group Structure S4 S3 S2 S2 S2 S3 S2 S2 S3 S2 S2 S2 S2 S3 S2 S2 S2 1 The subgroups {S2x S2} give additional structure … Observed Bifurcation Structure q* Group Structure S4 12, 34 13, 24 14, 23 The subgroups {S2x S2} give additional structure … Observed Bifurcation Structure q* Group Structure S4 12, 34 13, 24 14, 23 Theorem: There are at exactly K bifurcations on the branch (q1/N , ) whenever G(q1/N) is nonsingular Observed Bifurcation Structure q* There are K=52 bifurcations on the first branch A partial subgroup lattice for S4 and the corresponding bifurcating directions given by the Equivariant Branching Lemma S4 S3 3v v v v 0 S2 S2 S2 0 2v v v 0 0 v 2v v 0 0 v v 2v 0 S3 v 3v v v 0 S2 S2 S2 2v 0 v v 0 v 0 2v v 0 v 0 v 2v 0 1 S3 v v 3v v 0 S2 S2 S2 2v v 0 v 0 v 2v 0 v 0 v v 0 2v 0 S3 v v v 3v 0 S2 S2 S2 2v v v 0 0 v 2v v 0 0 v v 2v 0 0 A partial subgroup lattice for S4 and the corresponding bifurcating directions corresponding to subgroups isomorphic to S2 x S2. S4 12, 34 v v v v 13, 24 14, 23 v v v v v v v v This theory enables us to answer the questions previously posed … ?????? Observed Bifurcations for the 4 Blob Problem Conceptual Bifurcation Structure q* Why are there only 3 bifurcations observed? In general, are there only N-1 bifurcations? What kinds of bifurcations do we expect: pitchfork-like, transcritical, saddle-node, or some other type? How many bifurcating solutions are there? What do the bifurcating branches look like? Are they subcritical or supercritical ? What is the stability of the bifurcating branches? Is there always a bifurcating branch which contains solutions of the optimization problem? Are there bifurcations after all of the classes have resolved ? Conceptual Bifurcation Structure S4 q* S3 S2 S2 S3 S2 S2 S2 S3 S2 S2 S2 S2 S3 S2 S2 S2 1 Why are there only 3 bifurcations observed? In general, are there only N-1 bifurcations? There are N-1 symmetry breaking bifurcations from SM to SM-1 for M N. What kinds of bifurcations do we expect: pitchfork-like, transcritical, saddle-node, or some other type? How many bifurcating solutions are there? There are at least N from the first bifurcation, at least N-1 from the next one, etc. What do the bifurcating branches look like? They are subcritical or supercritical depending on the sign of the bifurcation discriminator (q*,*,uk) . What is the stability of the bifurcating branches? Is there always a bifurcating branch which contains solutions of the optimization problem? No. Are there bifurcations after all of the classes have resolved ? Generically, no. Continuation techniques numerically illustrate the theory using the Information Distortion I(Y,Z) bits q* Bifurcating branches with symmetry S2 x S2 = <(12),(34)> I(Y,Z) bits q* I(Y,Z) bits Additional structure!! I(Y,Z) bits I(Y,Z) bits A closer look … q* I(Y,Z) bits Bifurcation from S4 to S3… q* I(Y,Z) bits I(Y,Z) bits The bifurcation from S4 to S3 is subcritical … (the theory predicted this since the bifurcation discriminator (q1/4,*,u)<0 ) I(Y,Z) bits What does this mean regarding solutions of the original problems? (4) RH(I0) = maxqH(Z|X) constrained by I(Y,Z) I0 (7) maxq(H(Z|X) + I(Y,Z)) Theorem: • dR/dI0 = -(I0) • d2R/dI02 = -d(I0)/dI0 (4) RH(I0) = maxqH(Z|X) constrained by I(Y,Z) I0 (7) maxq(H(Z|X) + I(Y,Z)) RH as a function of I0 RH(I0) = maxq H(Z|Y) constrained by I(X;Z) I0 • is not convex and not concave • is a monotonically decreasing, continuous function RH Theorem: • dR/dI0 = -(I0) • d2R/dI02 = d(I0)/dI0 Consequences?? • Analogue for the Information Distortion RH(I0) = maxq H(Z|X) constrained by I(Y;Z) I0 is neither concave nor convex since subcritical bifurcations and saddle nodes exist. • Rate Distortion Function (from Information Theory) R(D0) = minq I(X;Z) constrained by D(X,Z) D0 is convex if D(Y,Z) is linear in q (Rose, 1994; Cover and Thomas; Grey). • Relevance Compression Function (for Information Bottleneck) RI(I0) = minq I(X;Z) constrained by I(Y;Z) I0 is convex if N>K+1 (Witsenhausen and Wyner 1975, Bachrach et al 2003) So What?? Analogue for the Information Distortion RH(I0) = maxq H(Z|X) constrained by I(Y;Z) I0 is neither concave nor convex since subcritical bifurcations and saddle nodes exist. Relevance Compression Function (for Information Bottleneck) RI(I0) = minq I(X;Z) constrained by I(Y;Z) I0 is convex if N>K+1 (Bachrach et al 2003) • RI(I0) and RH(I0) are related by I(X;Z) = H(Z) - H(Z|X). • The Information Bottleneck can not have a subcritical bifurcation when N > K+1. Are there subcritical bifurcations when N<K+1 ? • Is RH(I0) convex when N>K+1 ? That would mean that the subcritical bifurcations go away when considering the gradient flow in R(K+2)K instead of RNK. Application to cricket sensory data E(Y|Z): stimulus means conditioned on each of the classes spike patterns optimal clustering Conclusions … We have a complete theoretical picture of how the clusterings evolve for a class of annealing problems of the form maxq(G(q)+D(q)) subject to the assumptions stated earlier. o When clustering to N classes, there are N-1 bifurcations. o In general, there are only pitchfork and saddle-node bifurcations. o We can determine whether pitchfork bifurcations are either subcritical or supercritical (1st or 2nd order phase transitions) o We know the explicit bifurcating directions SO WHAT?? There are theoretical consequences … This suggests an algorithm for solving the annealing problem … (NIPS 2002)