User:Kiefer.Wolfowitz/Sandbox/SFS

Shapley–Folkman lemma.svg, the Shapley–Folkman lemma and the Shapley–Folkman–Starr theorem study the Minkowski addition of sets in a vector space. Minkowski addition is defined by the addition of the sets' members; for example, adding the set consisting of the integers zero and one to itself yields the set of the integers zero, one, and two:
 * The Shapley–Folkman lemma is illustrated by the Minkowski addition of four sets. The point (+) in the convex hull of the Minkowski sum of the four non-convex sets (right) is the sum of four points (+) from the (left-hand) sets—two points in two non-convex sets plus two points in the convex hulls of two sets. The convex hulls are shaded pink. The original sets each have exactly two points (shown as red dots). ]]
 * {0, 1} + {0, 1} = {0+0, 0+1, 1+0, 1+1} = {0, 1, 2}.

The Shapley–Folkman–Starr results

address the question, "Is the sum of many sets close to being convex?" A set is defined to be convex if every line segment joining two of its points is a subset in the set: For example, a solid disk $$\bullet$$ is convex but a circle $$\circ$$ is not, because the line segment joining two distinct points $$\oslash$$ is not a subset of the circle. The Shapley–Folkman–Starr results suggest that if the number of summed sets exceeds the dimension of the vector space, then their Minkowski sum is approximately convex.

The Shapley–Folkman–Starr theorem states an upper bound on the distance between the Minkowski sum and its convex hull—the convex hull of the Minkowski sum is the smallest convex set that contains the Minkowski sum. This distance is zero exactly when the sum is convex. Their bound on the distance depends on the dimension D and on the shapes of the summand-sets, but not on the number of summand-sets N, when N > D. The shapes of a subcollection of only D summand-sets determine the bound on the distance between the average sumset and its convex hull; thus, as the number of summands increases to infinity, the bound decreases to zero (for summand-sets of uniformly bounded size).

The lemma of Lloyd Shapley and Jon Folkman was first published by the economist Ross M. Starr, who was investigating the existence of economic equilibria that do not require consumer preferences to be convex. In his paper, Starr proved that a "convexified" economy has equilibria that are closely approximated by "quasi-equilibria" of the original economy; moreover, he proved that every quasi-equilbrium has many of the optimal properties of true equilibria, which are proved to exist for convex economies. Following Starr's 1969 paper, the Shapley–Folkman–Starr results have been widely used to show that central results of (convex) economic theory are good approximations to large economies with non-convexities. "The derivation of these results in general form has been one of the major achievements of postwar economic theory", wrote Roger Guesnerie.

The Shapley–Folkman lemma has been applied also to optimization and probability theory. In optimization theory, the Shapley–Folkman lemma has been used to explain the solution of minimization problems that are sums of many functions. The Shapley–Folkman lemma has been used also in proofs of the "law of averages" for random sets, a theorem that had been proved for only convex sets.

Introductory example
For example, the subset of the integers {0, 1, 2} is contained in the interval of real numbers [0, 2], which is convex. The Shapley–Folkman lemma implies that every point in [0, 2] is the sum of an integer from {0, 1} and a real number from [0, 1].

The distance between the convex interval [0, 2] and the non-convex set {0, 1, 2} equals one-half
 * 1/2 = |1-1/2| = |0-1/2| = |2-3/2| = |1-3/2|.

However, the distance between the averaged set
 * 1/2 ( {0, 1} + {0, 1} ) = {0, 1/2, 1}

and its convex hull [0, 1] is only 1/4, which is half the distance (1/2) between its summand {0, 1} and [0, 1]. As more sets are added and their sum is averaged, the average sumset "fills out" its convex hull: The maximum distance between the average sumset and its convex hull approaches zero as additional sets are averaged.

Preliminaries
The Shapley–Folkman lemma depends upon the following definitions and results from convex geometry.

Real vector spaces
A real vector space of two dimensions can be given a Cartesian coordinate system in which every point is identified by a list of two real numbers, called "coordinates", which are conventionally denoted by x and y. Two points in the Cartesian plane can be added coordinate-wise
 * (x1, y1) + (x2, y2) = (x1+x2, y1+y2);

further, a point can be multiplied by each real number λ coordinate-wise
 * λ (x, y) = (λx, λy).

More generally, any real vector space of (finite) dimension D can be viewed as the set of all possible lists of D real numbers {{nowrap|{ (v1, v2, . . ., v{{sub|D}})}} } together with two operations: vector addition and multiplication by a real number. For finite-dimensional vector spaces, the operations of vector addition and real-number multiplication can each be defined coordinate-wise, following the example of the Cartesian plane.

Convex sets
In a real vector space, a non-empty set Q is defined to be convex if, for each pair of its points, every point on the line segment that joins them is a subset of Q. For example, a solid disk $$\bullet$$ is convex but a circle $$\circ$$ is not, because it does not contain a line segment joining its points $$\oslash$$; the non-convex set of three integers {0, 1, 2} is contained in the interval [0, 2], which is convex. For example, a solid cube is convex; however, anything that is hollow or dented, for example, a crescent shape, is non-convex. The empty set is convex, either by definition or vacuously, depending on the author.

More formally, a set Q is convex if, for all points v0 and v1 in Q and for every real number λ in the unit interval [0,1], the point
 * (1 &minus; λ) v0 + λv1

is a member of Q.

By mathematical induction, a set Q is convex if and only if every convex combination of members of Q also belongs to Q. By definition, a convex combination of an indexed subset {v0, v1,. . ., vD} of a vector space is any weighted average λ0v0 + λ1v1 +. . . + λDvD, for some indexed set of non-negative real numbers {λd} satisfying the equation λ0 + λ1 +. . . + λD = 1.

The definition of a convex set implies that the intersection of two convex sets is a convex set. More generally, the intersection of a family of convex sets is a convex set. In particular, the intersection of two disjoint sets is the empty set, which is convex.

Convex hull
For every subset Q of a real vector space, its convex hull Conv(Q) is the minimal convex set that contains Q. Thus Conv(Q) is the intersection of all the convex sets that cover Q. The convex hull of a set can be equivalently defined to be the set of all convex combinations of points in Q. For example, the convex hull of the set of integers {0,1} is the closed interval of real numbers [0,1], which contains the integer end-points. The convex hull of the unit circle is the closed unit disk, which contains the unit circle.

Minkowski addition
In a real vector space, the Minkowski sum of two (non-empty) sets Q1 and Q2 is defined to be the set Q1 + Q2 formed by the addition of vectors element-wise from the summand sets
 * Q1 + Q2 = { q1 + q2 : q1 ∈ Q1 and q2 ∈ Q2 }.

[[File:Shapley–Folkman lemma.svg|thumb|300px|alt=The Shapley–Folkman lemma depicted by a diagram with two panes, one on the left and the other on the right. The left-hand pane displays four sets, which are displayed in a two-by-two array. Each of the sets contains exactly two points, which are displayed in red. In each set, the two points are joined by a pink line-segment, which is the convex hull of the original set. Each set has exactly one point that is indicated with a plus-symbol. In the top row of the two-by-two array, the plus-symbol lies in the interior of the line segment; in the bottom row, the plus-symbol coincides with one of the red-points. This completes the description of the left-hand pane of the diagram. The right-hand pane displays the Minkowski sum of the sets, which is the union of the sums having exactly one point from each summand-set; for the displayed sets, the sixteen sums are distinct points, which are displayed in red: The right-hand red sum-points are the sums of the left-hand red summand-points. The convex hull of the sixteen red-points is shaded in pink. In the pink interior of the right-hand sumset lies exactly one plus-symbol, which is the (unique) sum of the plus-symbols from the right-hand side. The right-hand plus-symbol is indeed the sum of the four plus-symbols from the left-hand sets, precisely two points from the original non-convex summand-sets and two points from the convex hulls of the remaining summand-sets.
 * Minkowski addition and convex hulls. The sixteen dark-red points (on the right) form the Minkowski sum of the four non-convex sets (on the left), each of which consists of a pair of red points. Their convex hulls (shaded pink) contain plus-signs (+): The right plus-sign is the sum of the left plus-signs.]]

For example
 * {0, 1} + {0, 1} = {0+0, 0+1, 1+0, 1+1} = {0, 1, 2}.

By the principle of mathematical induction, the Minkowski sum of a finite family of (non-empty) sets
 * {Qn : Qn ≠ Ø and 1 ≤ n ≤ N }

is the set formed by element-wise addition of vectors
 * ∑ Qn = {∑ qn : qn ∈ Qn}.

Convex hulls of Minkowski sums
Minkowski addition behaves well with respect to "convexification"—the operation of taking convex hulls. Specifically, for all subsets Q1 and Q2 of a real vector space, the convex hull of their Minkowski sum is the Minkowski sum of their convex hulls. That is,
 * Conv( Q1 + Q2 ) = Conv( Q1 ) + Conv( Q2 ).

This result holds more generally, as a consequence of the principle of mathematical induction. For each finite collection of sets,
 * Conv( ∑ Qn  ) = ∑ Conv( Qn ).

Statements
The preceding identity Conv( ∑ Qn ) = ∑ Conv( Qn ) implies that if a point x lies in the convex hull of the Minkowski sum of N sets
 * x ∈ Conv( ∑ Qn )

then x lies in the sum of the convex hulls of the summand-sets
 * x ∈ ∑ Conv( Qn ).

By the definition of Minkowski addition, this last expression means that x = ∑ qn for some selection of points qn in the convex hulls of the summand-sets, that is, where each qn ∈ Conv(Qn). In this representation, the selection of the summand-points qn depends on the chosen sum-point x.

Lemma of Shapley and Folkman
For this representation of the point x, the Shapley–Folkman lemma states that if the dimension D is less than the number of summands
 * D &lt; N

then convexification is needed for only D summand-sets, whose choice depends on x: The point has a representation where qd belongs to the convex hull of Qd for D (or fewer) summand-sets and qn belongs to Qn itself for the remaining sets. That is,
 * $$ x = \sum_{1\leq{d}\leq{D}}{q_d} + \sum_{D+1\leq{n}\leq{N}}{q_n} $$
 * $$ x \in{ \sum_{1\leq{d}\leq{D}}{\operatorname{Conv}{(Q_d)}} + \sum_{D+1\leq{n}\leq{N}}{Q_n} }$$

for some re-indexing of the summand sets; this re-indexing depends on the particular point x being represented.

The Shapley–Folkman lemma implies that every point in [0, 2] is the sum of an integer from {0, 1} and a real number from [0, 1].

Conversely, the Shapley–Folkman lemma characterizes the dimension of finite-dimensional, real vector spaces. That is, if a vector space obeys the Shapley–Folkman lemma for a natural number D, and for no number lesser than D, then its dimension is exactly D; the Shapley–Folkman lemma holds for only finite-dimensional vector spaces.

Shapley–Folkman theorem and Starr's corollary


Shapley and Folkman used their lemma to prove their  theorem: The Shapley–Folkman theorem states a bound on the distance between the Minkowski sum and its convex hull; this distance is zero exactly when the sum is convex. Their bound on the distance depends on the dimension D and on the shapes of the summand-sets, but not on the number of summand-sets N, when N > D.
 * The Shapley–Folkman theorem states that the squared Euclidean distance from any point in the convexified sum Conv( ∑ Qn ) to the original (unconvexified) sum ∑ Qn is bounded by the sum of the squares of the D largest circumradii of the sets Qn (the radii of the smallest spheres enclosing these sets). This bound is independent of the number of summand-sets N (if N &gt; D).

The circumradius often exceeds (and cannot be less than) the inner radius:

Starr used the inner radius to strengthen the conclusion of the Shapley–Folkman theorem: Starr's corollary states an upper bound on the Euclidean distance  between the Minkowski sum of N sets and the convex hull of the Minkowski sum; this distance between the sumset and its convex hull is a measurement of the non-convexity of the set. For simplicity, this distance is called the "non-convexity" of the set (with respect to Starr's measurement). Thus, Starr's bound on the non-convexity of the sumset depends on only the D largest inner radii of the summand-sets; however, Starr's bound does not depend on the number of summand-sets N, when N &gt; D. For example, the distance between the convex interval [0, 2] and the non-convex set {0, 1, 2} equals one-half
 * The inner radius of a set Qn is defined to be the smallest number r such that, for any point q in the convex hull of Qn, there is a sphere of radius r that contains a subset of Qn whose convex hull contains x.
 * Starr's corollary to the Shapley–Folkman theorem states that the squared Euclidean distance from any point x in the convexified sum Conv( ∑ Qn ) to the original (unconvexified) sum ∑ Qn is bounded by the sum of the squares of the D largest inner-radii of the sets Qn.
 * 1/2 = |1-1/2| = |0-1/2| = |2-3/2| = |1-3/2|.

Thus, Starr's bound on the non-convexity of the average sumset
 * $1/N$ ∑ Qn

decreases as the number of summands N increases. For example, the distance between the averaged set
 * 1/2 ( {0, 1} + {0, 1} ) = {0, 1/2, 1}

and its convex hull [0, 1] is only 1/4, which is half the distance (1/2) between its summand {0, 1} and [0, 1]. The shapes of a subcollection of only D summand-sets determine the bound on the distance between the average sumset and its convex hull; thus, as the number of summands increases to infinity, the bound decreases to zero (for summand-sets of uniformly bounded size). In fact, Starr's bound on the non-convexity of this average sumset decreases to zero as the number of summands N increases to infinity (when the inner radii of all the summands are bounded by the same number).

Proofs and computations
The original proof of the Shapley–Folkman lemma established only the existence of the representation, but did not provide an algorithm for computing the representation: Similar proofs have been given by Arrow and Hahn, Cassels, and Schneider, among others. An abstract and elegant proof by Ekeland has been extended by Artstein. Different proofs have appeared in unpublished papers, also. In 1981, Starr published an iterative method for computing a representation of a given sum-point; however, his computational proof provides a weaker bound than does the original result.

Applications
The Shapley–Folkman lemma has applications in economics, in optimization theory, and in probability.

Economics


In economics, a consumer's preferences are defined over all "baskets" of goods. Each basket is represented as a non-negative vector, whose coordinates represent the quantities of the goods. On this set of baskets, an indifference curve is defined for each consumer; a consumer's indifference curve contains all the baskets of commodities that the consumer regards as equivalent: That is, for every pair of baskets on the same indifference curve, the consumer does not prefer one basket over another. Through each basket of commodities passes one indifference curve. A consumer's preference set (relative to an indifference curve) is the union of the indifference curve and all the commodity baskets preferred by the consumer. A consumer's preferences are convex if all such preference sets are convex.

An optimal basket of goods occurs where the budget-line supports a consumer's preference set, as shown in the diagram. The set of optimal baskets is called the consumer's demand, which is a function of the prices. If the preference set is convex, then at every price the consumer's demand is a convex set, for example, a unique optimal basket or a line-segment of baskets.

Non-convex preferences
However, if a preference set is non-convex, then some prices determine a budget-line that supports two separate optimal-baskets. For example, we can imagine that, for zoos, a lion costs as much as an eagle, and further that a zoo's budget suffices for one eagle or one lion. We can suppose also that a zoo-keeper views either animal as equally valuable. In this case, the zoo would purchase either one lion or one eagle. Of course, a contemporary zoo-keeper does not want to purchase half of an eagle and half of a lion (or a griffin)! Thus, the zoo-keeper's preferences are non-convex: The zoo-keeper prefers having either animal to having any strictly convex combination of both.

When the consumer's preference set is non-convex, then (for some prices) the consumer's demand is not connected; a disconnected demand implies some discontinuous behavior by the consumer, as discussed by Harold Hotelling: If indifference curves for purchases be thought of as possessing a wavy character, convex to the origin in some regions and concave in others, we are forced to the conclusion that it is only the portions convex to the origin that can be regarded as possessing any importance, since the others are essentially unobservable. They can be detected only by the discontinuities that may occur in demand with variation in price-ratios, leading to an abrupt jumping of a point of tangency across a chasm when the straight line is rotated. But, while such discontinuities may reveal the existence of chasms, they can never measure their depth. The concave portions of the indifference curves and their many-dimensional generalizations, if they exist, must forever remain in unmeasurable obscurity.

The difficulties of studying non-convex preferences were emphasized by Herman Wold and again by Paul Samuelson, who wrote that non-convexities are "shrouded in eternal darkness ...", according to Diewert.

Nonetheless, non-convex preferences were illuminated from 1959 to 1961 by a sequence of papers in The Journal of Political Economy (JPE). The main contributors were Farrell, Bator, Koopmans,  and Rothenberg. In particular, Rothenberg's paper discussed the approximate convexity of sums of non-convex sets. These JPE-papers stimulated a paper by Lloyd Shapley and Martin Shubik, which considered convexified consumer-preferences and introduced the concept of an "approximate equilibrium". The JPE-papers and the Shapley–Shubik paper influenced another notion of "quasi-equilibria", due to Robert Aumann.

Starr's 1969 paper and contemporary economics
The previously noted papers were listed in an annotated bibliography that Kenneth Arrow gave Starr, who was then an undergraduate enrolled in Arrow's (graduate) advanced mathematical-economics course. In his term-paper, Starr studied the general equilibria of an artificial economy in which non-convex preferences were replaced by their convex hulls. In the convexified economy, at each price, the aggregate demand was the sum of convex hulls of the consumers' demands. Starr's ideas interested the mathematicians Lloyd Shapley and Jon Folkman, who proved their eponymous lemma and  theorem in "private correspondence",  which was reported by Starr's published paper of 1969.

In his 1969 publication, Starr applied the Shapley–Folkman–Starr theorem. Starr proved that the "convexified" economy has general equilibria that can be closely approximated by "quasi-equilbria" of the original economy, when the number of agents exceeds the dimension of the goods: Concretely, Starr proved that there exists there exists at least one quasi-equilibrium of prices popt with the following properties:


 * For each quasi-equilibrium's prices popt, all consumers can choose optimal baskets (maximally preferred and meeting their budget constraints).


 * At quasi-equilibrium prices popt in the convexified economy, every good's market is in equilibrium: Its supply equals its demand.


 * For each quasi-equilibrium, the prices "nearly clear" the markets for the original economy: an upper bound on the distance between the set of equilibria of the "convexified" economy and the set of quasi-equilibria of the original economy followed from Starr's corollary to the Shapley–Folkman theorem.

Starr established that "in the aggregate, the discrepancy between an allocation in the fictitious economy generated by [taking the convex hulls of all of the consumption and production sets] and some allocation in the real economy is bounded in a way that is independent of the number of economic agents. Therefore, the average agent experiences a deviation from intended actions that vanishes in significance as the number of agents goes to infinity". Following Starr's 1969 paper, the Shapley–Folkman–Starr results have been widely used in economic theory. Roger Guesnerie summarized their economic implications: "Some key results obtained under the convexity assumption remain (approximately) relevant in circumstances where convexity fails. For example, in economies with a large consumption side, preference nonconvexities do not destroy the standard results". "The derivation of these results in general form has been one of the major achievements of postwar economic theory", wrote Guesnerie. The Shapley–Folkman–Starr results have been featured in the economics literature: in microeconomics, in general-equilibrium theory, in public economics (including market failures), as well as in game theory, in mathematical economics, and in applied mathematics (for economists). The Shapley–Folkman–Starr results have also influenced economics research using measure and integration theory.

Mathematical optimization
The Shapley–Folkman lemma has been used to explain why large minimization problems with non-convexities can be nearly solved (with iterative methods whose convergence proofs are stated for only convex problems).

Preliminaries of optimization theory
Nonlinear optimization relies on the following definitions for real-valued functions:


 * The graph of a function f is the set of the pairs of arguments x and function evaluations f(x)
 * Graph(f) = {   ( x, f(x) )   }


 * The epigraph of a function f is the set of points above the graph
 * Epi(f) = { (x, u) : f(x) ≤ u }.


 * A real-valued function is defined to be a convex function if its epigraph is a convex set.

For example, the quadratic function f(x) = x2 is convex, as is the absolute value function g(x) = |x|. However, the sine function (pictured) is non-convex on the interval (0, π).

Additive optimization problems
In many optimization problems, the objective function f is separable: that is, f is the sum of many summand-functions, each of which has its its own argument:


 * f(x) = f ( (x1, ..., xN) ) = ∑ fn(xn).

For example, problems of linear optimization are separable. Given a separable problem with an optimal solution, we fix an optimal solution


 * xmin = (x1, ..., xN)min

with the minimum value f(xmin). For this separable problem, we also consider an optimal solution ( xmin, f(xmin) ) to the "convexified problem", where convex hulls are taken of the graphs of the summand functions. Such an optimal solution is the limit of a sequence of points in the convexified problem An application of the Shapley–Folkman lemma represents the given optimal-point as a sum of points in the graphs of the original summands and of a small number of convexified summands.
 * ( xj, f(xj) ) ∈  ∑ Conv ( Graph( fn ) ).

This analysis was published by Ivar Ekeland in 1974 to explain the apparent convexity of separable problems with many summands, despite the non-convexity of the summand problems. In 1973, the young mathematician Claude Lemaréchal was surprised by his success with convex minimization methods on problems that were known to be non-convex. Ekeland's analysis explained the success of methods of convex minimization on large and separable problems, despite the non-convexities of the summand functions. The Shapley–Folkman lemma has encouraged the use of methods of convex minimization on other applications with sums of many functions.

Probability and measure theory
Convex sets are often studied with probability theory. Each point in the convex hull of a (non-empty) subset of a finite-dimensional space is the expected value of a simple random vector that takes its values in a subset (by Carathéodory's lemma). This correspondence between convex sets and simple random vectors implies that the Shapley–Folkman–Starr results are useful in probability theory. In the other direction, probability theory provides tools to examine convex sets generally and the Shapley–Folkman–Starr results specifically. The Shapley–Folkman–Starr results have been widely used in the probabilistic theory of random sets, for example, to prove a law of large numbers, a central limit theorem, and a large-deviations principle. These proofs of probabilistic limit theorems used the Shapley–Folkman–Starr results to avoid the assumption that all the random sets be convex.

In measure theory, the Shapley–Folkman lemma enables a refinement of the Brunn–Minkowski inequality, which bounds the volume of sumsets in terms of the volumes of their summand-sets. The volume of a set is defined in terms of the Lebesgue measure, which is defined on subsets of Euclidean space. In advanced measure-theory, the Shapley–Folkman lemma has been used to prove Lyapunov's theorem, which states that the range of a vector measure is convex. Here, the traditional term "range" (alternatively, "image") is the set of values produced by the function. A vector measure is a vector-valued generalization of a measure; for example, if p1 and p2 are probability measures defined on the same probability space, then the function (p1, p2) is a vector measure. Lyapunov's theorem has been used in economics, in ("bang-bang") control theory, and in statistical theory. Lyapunov's theorem has been called a continuous counterpart of the Shapley–Folkman lemma, which has itself been called a discrete analogue of Lyapunov's theorem.