A VARIATIONAL METHOD FOR THE ANALYSIS OF A MONOTONE SCHEME FOR THE MONGE-AMPÈRE EQUATION 1. INTRODUCTION

Size: px
Start display at page:

Download "A VARIATIONAL METHOD FOR THE ANALYSIS OF A MONOTONE SCHEME FOR THE MONGE-AMPÈRE EQUATION 1. INTRODUCTION"

Transcription

1 A VARIATIONAL METHOD FOR THE ANALYSIS OF A MONOTONE SCHEME FOR THE MONGE-AMPÈRE EQUATION GERARD AWANOU AND LEOPOLD MATAMBA MESSI ABSTRACT. We give a proof of existence of a solution to the discrete problem obtained from a discretization of the Monge-Ampère equation by a monotone scheme. We prove that the unique minimizer of a convex functional of the gradient under convexity and nonlinear constraints solves the finite difference equation. The variational method with the monotone scheme gives a numerical method which is shown to be numerically stable when the right hand side of the equation is a linear combination of Dirac masses, i.e. for Aleksandrov solutions. 1. INTRODUCTION We are interested in the numerical resolution of the Monge-Ampère equation (1.1) det 2 u = f in Ω, u = g on Ω, where the domain Ω R n is assumed to be bounded and convex with boundary Ω. For a smooth function u, 2 u is the Hessian of u with entries ( 2 u)/( x i x j ), i, j = 1,..., n. In general the expression det 2 u should be interpreted in the sense of Aleksandrov, c.f. Section 4. The functions f and g are given with f 0. We make the assumption that f L 1 (Ω) and g C 0,1 ( Ω) can be extended to a function g C(Ω) which is convex on Ω. Under these assumptions, (1.1) is known to have a unique convex Aleksandrov solution [11]. To discretize (1.1) one needs a notion of discrete convexity as well as a discretization of det 2 u. As explained in [4, Section 1.2], it has been an open problem to prove the existence of a solution to the discrete problem obtained when one uses the monotone scheme of [7] and a notion of discrete convexity, also used in [7], which we will refer to as wide stencil convexity. Our strategy is based on a remark of P.L. Lions [13] which asserts that the unique C 0,1 (Ω) solution of (1.1) in the sense of Aleksandrov is the unique minimizer of J(w) = Φ( w) dx, over the convex set (1.2) Ω S = { v C(Ω), v = g on Ω, v convex on Ω, (det 2 v) 1 n f 1 n in Ω }, where Φ is a nonnegative strictly convex function satisfying (2.1) below. Mimicking the above principle at the discrete level, we prove the existence of a unique minimizer to the discrete optimization problem and shows that the unique minimizer solves the finite difference equations obtained with the monotone discretization. The notation C 0,1 (Ω) stands for the set of Lipschitz continuous functions on Ω. We present numerical evidence of stability of the discretization with a test case for which the right hand side is a linear combination of Dirac masses. We do not address in this paper the convergence of the discretization proposed in [7] to the Aleksandrov solution and refer to [5] for an axiomatic This research was supported in part by NSF grant DMS to Gerard Awanou and NSF grant DMS to the Mathematical Biosciences Institute. 1

2 2 GERARD AWANOU AND LEOPOLD MATAMBA MESSI approach to the convergence of finite difference schemes to the Aleksandrov solution. The existence of a solution to the discrete problem is proved in this paper, while the other assumptions required in [5], except stability, are proved in [2]. Organization of the paper. The paper is organized as follows: in the next section, we introduce some notation and prove that when the equation is discretized with the monotone scheme of [7], the unique minimizer of the discrete problem solves the finite difference equation. Thus the latter has at least one solution. Section 3 is devoted to numerical results. In Section 4 we review the notion of Aleksandrov solution and explicitly compute the Monge-Ampère measure for a test function often used to evaluate the efficiency of numerical methods for Aleksandrov solutions [6]. 2. DISCRETE OPTIMIZATION PROBLEM We start with some notations needed to state the discrete optimization problem. All the functions we consider take finite values. We make the usual convention of denoting constants by the letter C. We make the following assumptions on the nonnegative convex function Φ (2.1a) Φ C 2 (R n ) (2.1b) Φ(p) C + C p (2.1c) ν > 0 such that Φ(p) Φ(q), p q ν p q 2, p, q R n where, denotes the Euclidean inner product in R n and. denotes the associated norm. Without loss of generality, we will assume, for the analysis in this paper, that Φ is strictly convex. For example, one may take Φ(p) = p 2. The results in [13] hold for more general convex functions Φ Notation and preliminaries. Most of the notation is taken from [1]. Let Ω = (0, 1) n, h = 1/N, N N and let M h denote the mesh which consists of points x = hz R n Ω, z Z n. Put M h = M h Ω, M h = M h Ω, and denote by U h the set of real valued functions defined on M h. Without loss of generality, for a function v defined on Ω, we use the notation v for the restriction of v to M h. We denote by e i, i = 1,..., n the i-th coordinate unit vector of R n and consider the first order difference operator defined on M h by v i h (x) := v h(x) v h (x he i ). h 2.2. Discretization of the functional J. We recall that for a convex function v on Ω, the one-sided partial derivatives are defined everywhere and denote by v the left-hand side gradient of v. Since the convex function v is differentiable almost everywhere, we have (2.2) J(w) = Φ( w) dx. Ω We now use the latter expression of J(w) to construct a discrete analogue of J(w) using Riemann sums. For x = (x 1,..., x n ) M h, we define P x = { y Ω, x i h y i x i }. It is easy to see that P x is nonempty if and only if either dist(x, Ω) < h n or dist(x h, Ω) < h n. Furthermore, we have x Mh P x = Ω and for x y, P x P y is a set of Lebesgue measure 0. For x M h, we denote by P x the Lebesgue measure of P x, i.e. P x = h n if P x Ω. We

3 use the notation x h/2 to denote the center of P x that is, the point obtained by subtracting h/2 from each coordinate x i of x. We define for v h U h and a convex real-valued function Φ on R n J h (v h ) = x M h h n Φ( h v h (x)), where h v h is the vector of backward finite differences of the mesh function v h, i.e. h v h (x) = ( i v h (x)),...,n. We assume that the sum in the definition of J h (v h ) is over the set of mesh points at which h v h (x) is defined. We note that for all i, i v h extends to a piecewise constant function on Ω, denoted also by i v h and taking the constant value i v h (x) on P x for x M h Monotone discretization of det 2 u and wide stencil convexity. We prove in this section that if one uses the monotone scheme introduced in [7], the discrete optimization problem has a unique minimizer which solves the corresponding finite difference equations. This also proves the existence of a solution to the latter. We do not know whether uniqueness holds for a finite difference system derived from a monotone scheme approximation of the Monge-Ampère operator. However the variational approach in this paper provides a method for selecting a specific solution of such a finite difference system. We define dθ(x) = max min v αh v =1 α h /x±α h M h α h. and let ζ 0 M h such that dθ(ζ 0 ) = min dθ(x). x M h The quantity dθ(ζ 0 ) has been referred to as directional resolution in [7]. Following [7, 3], we define a monotone Monge-Ampère operator by n v h (x + α i ) 2v h (x) + v h (x α i ) (2.3) M[v h ](x) = inf α i 2 for x M h, (α 1,...,α n) W h where W h denotes the set of orthogonal bases of R n such that for (α 1,..., α n ) W h, ζ 0 ± α i M h, i. For x M h, such that x + α i M h but x α i / M h for some i, we denote by x αi the intersection with Ω of the line through x and x + α i closest to x. We then define v h (x α i ) to be the value obtained by quadratic interpolation of v h (x αi ), v h (x) and v h (x + α i ). Similarly, if x α i M h but x + α i / M h for some i, we define v h (x + α i ) by quadratic interpolation of v h (x αi ), v h (x) and v h (x α i ), where now x αi is the intersection with Ω of the line through x and x α i closest to x. If both x α i and x + α i are not in M h, we use as interpolation points x and the intersection points with the boundary of the line through x α i and x + α i. For x M h and e Zn such that e is an element of an orthogonal base (α 1,..., α n ) W h, we define e v h (x) = v h (x + e) 2v h (x) + v h (x e) with the above convention. Definition 2.1. We say that a mesh function v h is wide stencil convex if and only if e v h (x) = v h (x + e) 2v h (x) + v h (x e) 0 for all x M h and e Zn for which e v h (x) is defined. Let C h denote the cone of wide stencil convex mesh functions. We recall that the discrete Laplacian takes the form d h v h (x) = hei v h (x), 3

4 4 GERARD AWANOU AND LEOPOLD MATAMBA MESSI where { e i, i = 1,..., d } denotes the canonical basis of R d. We define (2.4) S h = { v h C h, v h = g on M h, and (M[v h ](x)) 1 n f(x) 1 n, x M h }. We seek a minimizer of J h over S h and consider the problem: find u h C h such that (2.5) M[u h ](x) = f(x), x M h u h = g on M h. Lemma 2.2. The operator v h M[v h ] is concave and thus the set S h is convex. Proof. We note that for λ 0 (M[λv h ](x)) 1 n = λ(m[vh ](x)) 1 n. It is therefore enough to prove that for v h, w h C h, we have (2.6) (M[v h + w h ](x)) 1 n (M[vh ](x)) 1 n + (M[wh ](x)) 1 n. Let (α 1,..., α n ) W h (x) be such that with M[v h + w h ](x) = n (λ i 1 + λ i 2) λ i 1 = v h(x + α i ) 2v h (x) + v h (x α i ) α i 2 and λ i 2 = w h(x + α i ) 2w h (x) + w h (x α i ) α i 2. Since v h, w h C h, λ i 1, λi 2 0 for all i. By Minkowski s determinant theorem ( n ) 1 ( (λ i 1 + λ i n n ) 1 ( n n ) 1 n 2) +, from which (2.6) follows. Lemma 2.3. The convex set S h is nonempty. λ i 1 λ i 2 Proof. For each y in M h, let q y be a quadratic function such that M[q y ](x) = f(y) for all x in M h. For example, we may define q y by q y (x) = 1 2 f(y) x 2. Put ˆf = y M q y. Since g is bounded on Ω, we can find a number K such that ˆf K g on h Ω. We define w h U h by We claim that w h S h. w h (x) = ˆf(x) K, x M h w h = g on M h. For x M h and (α 1,..., α n ) W h such that x ± α i M h for all i, either w h (x + α i ) = ˆf(x+α i ) K or w h (x+α i ) = g(x+α i ) ˆf(x+α i ) K. Similarly w h (x α i ) ˆf(x α i ) K. We conclude that w h (x + α i ) 2w h (x) + w h (x α i ) ˆf(x + α i ) 2 ˆf(x) + ˆf(x α i ) α i 2 f(x). The case when w h (x + α i ) or w h (x α i ) is obtained by quadratic interpolation is similar. The quadratic function passing through the points (x α i, w h (x α i )), (x, w h (x)) and (x αi, g(x αi )) is above the corresponding section of the graph of ˆf(x) K and thus w h (x + α i ) ˆf(x + α i ) K. A similar argument gives w h (x α i ) ˆf(x α i ) K and we conclude as above.

5 5 Thus w h C h and (M[w h ]) 1/n f. We can now state the following result. Theorem 2.4. For Φ(p) = p 2, the functional J h has a unique minimizer u h in S h and u h solves the finite difference equation (2.5). Proof. Since Φ is strictly convex and S h is nonempty, it follows that the functional J h has a unique minimizer u h on the convex set S h. We now show that u h solves the finite difference system (2.5). To this end, it suffices to show that M[u h ] = f h on M h. Let us assume to the contrary that there exists x 0 M h such that M[u h ](x 0 ) > f h (x 0 ) 0. We have (2.7) 0 f h (x 0 ) < M[u h ](x 0 ) n u h (x 0 + α i ) 2u h (x 0 ) + u h (x 0 α i ) α i 2, for all (α 1,..., α n ) W h. Using the geometric mean inequality, we get f h (x 0 ) 1 n < (M[uh ](x 0 )) 1 1 αi u h (x 0 ) n n α i 2. We conclude that for all e Z n such that e u h (x 0 ) is defined, we have e u h (x 0 ) > 0. Let ɛ 0 = inf{ e u h (x 0 ), e Z n }. The mapping E : r n (u h(x 0 + α i ) + u h (x 0 α i ) 2r)/ α i 2 is continuous and by (2.7), with r 0 = u h (x 0 ), E(r 0 ) > f h (x 0 ). Therefore there exists ɛ 1 > 0 such that for r r 0 < ɛ 1, we have E(r) > f h (x 0 ). Finally, put ɛ = min(ɛ 0, ɛ 1 ). Following a strategy used in [12], we define w h by w h (x) = u h (x), x x 0, w h (x 0 ) = u h (x 0 ) + ɛ 4. By construction w h = g h on M h. For x x 0 either e w h (x) = e u h (x) or e w h (x) = e u h (x) + ɛ/4. Moreover e w h (x 0 ) = e u h (x 0 ) ɛ/2 ɛ 0 ɛ/2 ɛ/2 > 0 by the definition of ɛ. We conclude that w h C h. For x x 0, M[w h ](x) M[u h ](x) f h (x). We have n αi w h (x 0 ) n M[w h ](x 0 ) = α i 2 = αi u h (x 0 ) ɛ 2 α i 2, for some element (α 1,..., α n ) W h. That is, M[w h ](x 0 ) = E(r 0 + ɛ/4) > f h (x 0 ). Thus M[w h ](x) f h (x) for all x M h.

6 6 GERARD AWANOU AND LEOPOLD MATAMBA MESSI It remains to show that J h (w h ) < J h (u h ). Let M x0 denote the subset of M h consisting in x 0 and the points x 0 + he j, j = 1,..., n at which h u h is defined. We have J h (w h ) = h n h u h (x) 2 + h n h w h (x 0 ) 2 + h w h (x 0 + he j ) 2 x/ M x0 j=1 J h (w h ) = h n h u h (x) 2 + h n 2 (w h (x 0 ) w h (x 0 he i )) 2 x/ M x0 + h n 2 (w h (x 0 + he j ) w h (x 0 + he j he i )) 2 = h n j=1 x/ M x0 h u h (x) 2 + h n 2 j=1 (w h (x 0 + he j ) w h (x 0 + he j he i )) 2 i j + h n 2 (w h (x 0 ) w h (x 0 he i )) 2 + (w h (x 0 + he i ) w h (x 0 )) 2. However (w h (x 0 ) w h (x 0 he i )) 2 + (w h (x 0 + he i ) w h (x 0 )) 2 = ( uh (x 0 ) u h (x 0 he i ) + ɛ ) 2 ( + uh (x 0 + he i ) u h (x 0 ) ɛ ) 2 = 4 4 Thus, by our choice of ɛ, (u h (x 0 ) u h (x 0 he i )) 2 + (u h (x 0 + he i ) u h (x 0 )) 2 + nɛ2 8 ɛ 2 hu h (x 0 ). J h (w h ) = J h (u h ) + nɛ2 8 ɛ 2 hu h (x 0 ) = J h (u h ) + ɛ 2 (nɛ 4 hu h (x 0 )) < J h (u h ), since e u h (x 0 ) ɛ 0 and thus h u h (x 0 )) nɛ 0 nɛ > nɛ/4. This contradicts the assumption that u h is a minimizer and concludes the proof. 3. NUMERICAL EXPERIMENTS In this section, we report numerical results for the variational framework studied in this paper. A natural approach for solving the nonlinear finite difference equations is to use Newton s method [7]. However, it has been reported in [6] that the latter is not effective for Test 6 below for which the right hand side is a linear combination of Dirac masses. To illustrate the numerical stability of the variational method studied, we present results for a wide range of classical test functions. Our numerical results indicate that the discretization proposed in [7] is stable for Aleksandrov solutions. This observation combined with the numerical results given in [5] indicate that it may be possible to solve the nonlinear finite difference equations effectively for Aleksandrov solutions. We refer to [6] for a convergent discretization for such solutions. Throughout this section, Ω denotes the square (0, 1) (0, 1) and we take Φ(p) = p 2.

7 3.1. The boundary value problems. We tested the framework developed in this paper on six examples of varying regularity and covering the full range of solutions of the Monge-Ampère equation Approximating classical and viscosity solutions. The first boundary value problem, admitting a classical smooth solution, is given by (3.1) f(x) = (1 + x 2 ) exp( x 2 ) and g(x) = exp( x 2 2 ). The second boundary value problem with data 2 (3.2) f(x) = (2 x 2 ) 2 and g(x) = 2 x 2 admits a viscosity solution with unbounded gradient. No exact solution is known for the third example with data (3.3) f(x) = 1 and g(x) = Approximating Aleksandrov solutions. The fourth example corresponds to the Monge-Ampère Dirichlet boundary value problem with data (3.4) f(x) = 0 and g(x) = x For the fifth example the right hand side is taken as a Dirac mass δ (1/2,1/2) which is approximated with the mesh functions { π if max{ x f h (x) = 4h , x } < h 0 otherwise. In that case the exact solution is given by u(x 1, x 2 ) = (x 1 1/2) 2 + (x 2 1/2) 2. For the last and sixth example, we solve the equation det 2 u = µ where the measure µ is given by (3.5) µ = π 2 δ ( 1 4, 1 2 ) + π 2 δ ( 3 4, 1 2 ). We show in the appendix that a solution to det 2 u = µ is the function x 2 { 1 2 if 1 4 < x 1 < 3 } 4 u(x) = min (x )2 + (x )2, (x 1 34 )2 + (x 2 12 )2 otherwise. The corresponding mesh functions approximating µ were taken as f h (x) = π/(8h 2 )1 Dh (1/4,1/2) + π/(8h 2 )1 Dh (3/4,1/2) with D r (x) denoting the open ball of center x and radius r in the maximum norm. For this test we experimented with various approximations of the Dirac masses with different levels of accuracy. In particular taking the closed ball in the definition of the above mesh function leads to divergence for the 17 point scheme Convex program. The convex program associated with the monotone scheme reads as follows arg min h 2 Φ( h u h (x)) x M h (3.6) subject to u h = g on M h M[uh ](x) f(x) x M h. 7

8 8 GERARD AWANOU AND LEOPOLD MATAMBA MESSI For our purposes in this section, we use the 9-point monotone scheme for which there are exactly two pairs of orthogonal directions for each interior grid point and the 17 point scheme with quadratic interpolation for which there are exactly six pairs of orthogonal directions for each interior grid point Implementation. It is not our goal in this section to develop or identify the most efficient method for solving (3.6). A plethora of methods and algorithms to solve this type of convex programs, including primal-dual and barrier methods, are readily available in the literature. Rather, we aim to leverage these algorithms to solve (3.6) and provide numerical evidence supporting the analysis done in this work. In particular, we use the convex optimization toolbox CVX [8, 9] to formulate the optimization problem in Matlab and solve it using the infeasible primal-dual interior point algorithm SDPT3 [14] Numerical evidence of convergence. We provide numerical evidence for the six boundary value problems defined early in this section. For Tests 3, 4, 5 and 6 we graph the numerical solutions in Figure 1. For Tests 1, 2, 5 and 6 we give the maximum errors using the 9 point scheme, Table 1 and the 17 point scheme Table 2. h Test 1 Test 2 Test 5 Test TABLE 1. Maximum error for the 9 point scheme. h Test 1 Test 2 Test 5 Test TABLE 2. Maximum error for the 17 point scheme. We note that the numerical solution is bounded independently of h. However for the two Dirac mass example, accuracy deteriorates with the 17 point scheme. The numerical error does decrease with h in that case. After all the numerical solution is an approximate minimizer and with more complicated constraints it seems more difficile to find a suitable discrete minimizer. For example the computing resources for solving the optimization problem for h = 2 6 with the 17 point scheme far exceed the ones required for the 9 point scheme although both problems have the same number of constraints. To further illustrate our point that the discretization considered is stable for Aleksandrov solutions, we repeat the numerical experiment for Test 6 with the standard finite difference discretization. It is a 9 point scheme which, for smooth solutions, is more accurate than the 9 point monotone scheme. One could have expected divergence for singular solutions. But we obtained accurate results for all

9 9 (A) Numerical solution for Test 3. (B) Numerical solution for Test 4. (C) Numerical solution for Test 5. (D) Numerical solution for Test 6. FIGURE 1. Solutions computed on a grid with the 9 point scheme. h 1/2 2 1/2 3 1/2 4 1/2 5 1/ TABLE 3. Maximum error for Dirac masses with standard discretization of the determinant of the Hessian. the test cases. Here we present the results for Test 6, c.f. Table 3, where we take as the right hand side f h (x) = 2/h 2 1 Bh/2 (1/4,1/2) + 2/h 2 1 Bh/2 (3/4,1/2) with B r (x) denoting the Euclidean open ball of center x and radius r. The inequality constraints are now taken as det(h[uh ](x)) f(x) for x M h, where H[u h ](x) is the matrix obtained by discretizing the Hessian with second order accurate central finite difference approximations. The code used to generate the results is available upon request. 4. APPENDIX In this section we review the geometric definition of the notion of Monge-Ampère measure and the notion of Aleksandrov solution. We refer to [10] for further details.

10 10 GERARD AWANOU AND LEOPOLD MATAMBA MESSI The normal mapping or subdifferential of a real valued function v defined on Ω, is a set-valued mapping v defined from Ω to the set of subsets of R d such that for any x 0 Ω, For a subset E Ω, we define v(x 0 ) = { q R n : v(x) v(x 0 ) + q (x x 0 ), for all x Ω }. v(e) = v(x). We denote by E the Lebesgue measure of E when E is measurable. The class is a Borel σ-algebra and the mapping x E S = { E Ω, v(e) is Lebesgue measurable }, M[v] : S R, M[v](E) = v(e), is a measure, finite on compact sets, called the Monge-Ampère measure associated with the function v. Definition 4.1. Let µ be a Borel measure on Ω. A convex function on Ω is an Aleksandrov solution of det 2 v = µ if M[v] = µ. We now show that the Monge-Ampère measure on Ω = (0, 1) (0, 1) associated with the function x 2 { 1 2 if 1 4 < x 1 < 3 } 4 u(x 1, x 2 ) = min (x )2 + (x )2, (x 1 34 )2 + (x 2 12 )2 otherwise, is given by M[u] = π 2 δ ( 1 4, 1 2 ) + π 2 δ ( 3 4, 1 ). 2 To this end it suffices to prove that for (x 1, x 2 ) not in { (1/4, 1/2), (3/4, 1/2) }, u(x 1, x 2 ) is contained in a set of measure zero and u(x 1, 1/2) is a set of measure π/2 for x 1 {1/4, 3/4}. Indeed it is easy to see that for (x 1, x 2 ) Ω with x 1 1/4 or x 1 3/4 or x 2 1/2, the function u is differentiable and its gradient u satisfies u(x 1, x 2 ) = 1. For those points, the normal mapping of u is always a subset of the unit circle, that is, u(x 1, x 2 ) {p R 2 : p = 1}, and it follows that u(x 1, x 2 ) = 0 for (x 1, x 2 ) Ω with x or x or x So it remains to compute u(x 1, x 2 ) for (x 1, x 2 ) Ω with x 1 = 1/4 or x 1 = 3/4 or x 2 = 1/2. Fix (x 1, x 2 ) Ω such that x 1 = 1/4 or x 1 = 3/4 and let (p 1, p 2 ) u(x 1, x 2 ). By definition of the subdifferential, we have u(y 1, y 2 ) x p 1(y 1 x 1 ) + p 2 (y 2 x 2 ) for 0 < y 1, y 2 < 1. Letting y 1 = x 1 in the above inequality yields y x p 2(y 2 x 2 ) for 0 < y 1 < 1, and it follows that [ 1, 1] if x 2 = 1 2 p 2 x 1 2 (x 2) = 1 if x 2 > if x 2 < 1 2. Consequently for x 1 = 1/4 or x 1 = 3/4 and x 2 1/2, the subdifferential u(x 1, x 2 ) is a subset of the negligible set {p R 2 : p 2 = 1}, and for x 1 = 1/4 or x 1 = 3/4 we have u(x 1, 1/2)

11 11 R [ 1, 1]. Next, we show that { u(x 1, 1 p R 2 2 ) = : p 1 and p 1 0 } if x 1 = 1 4 { p R 2 : p 1 and p 1 0 } if x 1 = 3 4. Indeed for p u(x 1, 1/2), we have (4.1) u(y 1, y 2 ) p 1 (y 1 x 1 ) + p 2 (y ) for all 0 < y 1, y 2 < 1. In particular for y 2 = 1/2, we get p 1 (y 1 x 1 ) 0 for all 1/4 < y 1 < 3/4 and it follows that p 1 0 for x 1 = 1/4 and p 1 0 for x 1 = 3/4. We now consider a line segment through (x 1, 1/2) and direction p and contained in Ω. It has parametric equations (x 1 + tp 1, 1/2 + tp 2 ) for t > 0 sufficiently small. Since u(x 1, 1/2) = 0, by definition of subdifferential we have u(x 1 + tp 1, tp 2) t p 2. On the other hand, when x 1 = 1/4 we have p 1 0 and u(x 1 + tp 1, 1/2 + tp 2 ) t p. Similarly when x 1 = 3/4 we have p 1 0 and u(x 1 + tp 1, 1/2 + tp 2 ) t p. In summary t p 2 u(x 1 + tp 1, tp 2) t p, and it follows that p 1. We have proved that { u(x 1, 1 p R 2 2 ) : p 1 and p 1 0 } if x 1 = 1 4 { p R 2 : p 1 and p 1 0 } if x 1 = 3 4. To conclude the proof, we verify that the reverse inclusion holds as well. We verify it for the case x 1 = 1/4, the argument for x 1 = 3/4 being analogous. Let p R 2 be such that p 1 and p 1 0. Under these assumptions on p, for (y 1, y 2 ) Ω and z = (1/4, 1/2) we have p 1 (y 1 1/4) + p 2 (y 2 1/2) p 2 (y 2 1/2) if y 1 1/4 and p 1 (y ) + p 2(y 2 1 ) = p (y z) p (y z) 2 (y z) = (y )2 + (y )2. It follows that p 1 (y ) + p 2(y 2 1 { y2 2 ) 1 2 u(y 1, y 2 ) if y y y u(y 1, y 2 ) if y 1 < 1 4. Thus the inequality (4.1) holds for all (y 1, y 2 ) Ω and as a result p u(1/4, 1/2). u(1/4, 1/2) and u(3/4, 1/2) have measure π/2. Thus REFERENCES [1] N. E. Aguilera and P. Morin. Approximating optimization problems over convex functions. Numer. Math., 111(1):1 34, [2] G. Awanou. Discrete Aleksandrov solutions of the Monge-Ampère equation. awanou/up.html, [3] G. Awanou. Convergence rate of a stable, monotone and consistent scheme for the Monge-Ampère equation. Symmetry, 8(4):18, [4] G. Awanou. On standard finite difference discretizations of the elliptic Monge-Ampère equation To appear in J Sci Comput doi: /s y. [5] G. Awanou and R. Awi. Convergence of finite difference schemes to the Aleksandrov solution of the Monge-Ampère equation. Acta Applicandae Mathematicae, 144(1):87 98, 2016.

12 12 GERARD AWANOU AND LEOPOLD MATAMBA MESSI [6] J.-D. Benamou and B. D. Froese. A viscosity framework for computing Pogorelov solutions of the Monge-Ampère equation [7] B. Froese and A. Oberman. Convergent finite difference solvers for viscosity solutions of the elliptic Monge-Ampère equation in dimensions two and higher. SIAM J. Numer. Anal., 49(4): , [8] M. Grant. Disciplined Convex Programming. PhD thesis, Stanford University, [9] M. Grant and S. Boyd. CVX: Matlab software for disciplined convex programming, version com/cvx, Mar [10] C. E. Gutiérrez. The Monge-Ampère equation. Progress in Nonlinear Differential Equations and their Applications, 44. Birkhäuser Boston Inc., Boston, MA, [11] D. Hartenstine. The Dirichlet problem for the Monge-Ampère equation in convex (but not strictly convex) domains. Electron. J. Differential Equations, pages No. 138, 9 pp. (electronic), [12] G. F. LAWLER, Weak convergence of a random walk in a random environment, Comm. Math. Phys., 87 (1982/83), pp [13] P.-L. Lions. Two remarks on Monge-Ampère equations. Ann. Mat. Pura Appl. (4), 142: (1986), [14] K. C. Toh, M. J. Todd, and R. H. Tütüncü. A MATLAB software package for semidefinite programming, version (2): , DEPARTMENT OF MATHEMATICS, STATISTICS, AND COMPUTER SCIENCE, M/C 249. UNIVERSITY OF ILLINOIS AT CHICAGO, CHICAGO, IL , USA address: awanou@uic.edu URL: awanou MATHEMATICAL BIOSCIENCES INSTITUTE, THE OHIO STATE UNIVERSITY, COLUMBUS, OH 43210, USA address: matambamessi.1@mbi.osu.edu URL:

ALEKSANDROV-TYPE ESTIMATES FOR A PARABOLIC MONGE-AMPÈRE EQUATION

ALEKSANDROV-TYPE ESTIMATES FOR A PARABOLIC MONGE-AMPÈRE EQUATION Electronic Journal of Differential Equations, Vol. 2005(2005), No. 11, pp. 1 8. ISSN: 1072-6691. URL: http://ejde.math.txstate.edu or http://ejde.math.unt.edu ftp ejde.math.txstate.edu (login: ftp) ALEKSANDROV-TYPE

More information

Fast convergent finite difference solvers for the elliptic Monge-Ampère equation

Fast convergent finite difference solvers for the elliptic Monge-Ampère equation Fast convergent finite difference solvers for the elliptic Monge-Ampère equation Adam Oberman Simon Fraser University BIRS February 17, 211 Joint work [O.] 28. Convergent scheme in two dim. Explicit solver.

More information

ESTIMATES FOR THE MONGE-AMPERE EQUATION

ESTIMATES FOR THE MONGE-AMPERE EQUATION GLOBAL W 2,p ESTIMATES FOR THE MONGE-AMPERE EQUATION O. SAVIN Abstract. We use a localization property of boundary sections for solutions to the Monge-Ampere equation obtain global W 2,p estimates under

More information

1. Introduction Boundary estimates for the second derivatives of the solution to the Dirichlet problem for the Monge-Ampere equation

1. Introduction Boundary estimates for the second derivatives of the solution to the Dirichlet problem for the Monge-Ampere equation POINTWISE C 2,α ESTIMATES AT THE BOUNDARY FOR THE MONGE-AMPERE EQUATION O. SAVIN Abstract. We prove a localization property of boundary sections for solutions to the Monge-Ampere equation. As a consequence

More information

PARTIAL REGULARITY OF BRENIER SOLUTIONS OF THE MONGE-AMPÈRE EQUATION

PARTIAL REGULARITY OF BRENIER SOLUTIONS OF THE MONGE-AMPÈRE EQUATION PARTIAL REGULARITY OF BRENIER SOLUTIONS OF THE MONGE-AMPÈRE EQUATION ALESSIO FIGALLI AND YOUNG-HEON KIM Abstract. Given Ω, Λ R n two bounded open sets, and f and g two probability densities concentrated

More information

REGULARITY RESULTS FOR THE EQUATION u 11 u 22 = Introduction

REGULARITY RESULTS FOR THE EQUATION u 11 u 22 = Introduction REGULARITY RESULTS FOR THE EQUATION u 11 u 22 = 1 CONNOR MOONEY AND OVIDIU SAVIN Abstract. We study the equation u 11 u 22 = 1 in R 2. Our results include an interior C 2 estimate, classical solvability

More information

A LOCALIZATION PROPERTY AT THE BOUNDARY FOR MONGE-AMPERE EQUATION

A LOCALIZATION PROPERTY AT THE BOUNDARY FOR MONGE-AMPERE EQUATION A LOCALIZATION PROPERTY AT THE BOUNDARY FOR MONGE-AMPERE EQUATION O. SAVIN. Introduction In this paper we study the geometry of the sections for solutions to the Monge- Ampere equation det D 2 u = f, u

More information

REGULARITY OF MONOTONE TRANSPORT MAPS BETWEEN UNBOUNDED DOMAINS

REGULARITY OF MONOTONE TRANSPORT MAPS BETWEEN UNBOUNDED DOMAINS REGULARITY OF MONOTONE TRANSPORT MAPS BETWEEN UNBOUNDED DOMAINS DARIO CORDERO-ERAUSQUIN AND ALESSIO FIGALLI A Luis A. Caffarelli en su 70 años, con amistad y admiración Abstract. The regularity of monotone

More information

AFFINE MAXIMAL HYPERSURFACES. Xu-Jia Wang. Centre for Mathematics and Its Applications The Australian National University

AFFINE MAXIMAL HYPERSURFACES. Xu-Jia Wang. Centre for Mathematics and Its Applications The Australian National University AFFINE MAXIMAL HYPERSURFACES Xu-Jia Wang Centre for Mathematics and Its Applications The Australian National University Abstract. This is a brief survey of recent works by Neil Trudinger and myself on

More information

Degenerate Monge-Ampère equations and the smoothness of the eigenfunction

Degenerate Monge-Ampère equations and the smoothness of the eigenfunction Degenerate Monge-Ampère equations and the smoothness of the eigenfunction Ovidiu Savin Columbia University November 2nd, 2015 Ovidiu Savin (Columbia University) Degenerate Monge-Ampère equations November

More information

GERARD AWANOU. i 1< <i k

GERARD AWANOU. i 1< <i k ITERATIVE METHODS FOR k-hessian EQUATIONS GERARD AWANOU Abstract. On a domain of the n-dimensional Euclidean space, and for an integer k = 1,..., n, the k-hessian equations are fully nonlinear elliptic

More information

SHARP BOUNDARY TRACE INEQUALITIES. 1. Introduction

SHARP BOUNDARY TRACE INEQUALITIES. 1. Introduction SHARP BOUNDARY TRACE INEQUALITIES GILES AUCHMUTY Abstract. This paper describes sharp inequalities for the trace of Sobolev functions on the boundary of a bounded region R N. The inequalities bound (semi-)norms

More information

Centre for Mathematics and Its Applications The Australian National University Canberra, ACT 0200 Australia. 1. Introduction

Centre for Mathematics and Its Applications The Australian National University Canberra, ACT 0200 Australia. 1. Introduction ON LOCALLY CONVEX HYPERSURFACES WITH BOUNDARY Neil S. Trudinger Xu-Jia Wang Centre for Mathematics and Its Applications The Australian National University Canberra, ACT 0200 Australia Abstract. In this

More information

Asymptotic behavior of infinity harmonic functions near an isolated singularity

Asymptotic behavior of infinity harmonic functions near an isolated singularity Asymptotic behavior of infinity harmonic functions near an isolated singularity Ovidiu Savin, Changyou Wang, Yifeng Yu Abstract In this paper, we prove if n 2 x 0 is an isolated singularity of a nonegative

More information

On the Local Convergence of Regula-falsi-type Method for Generalized Equations

On the Local Convergence of Regula-falsi-type Method for Generalized Equations Journal of Advances in Applied Mathematics, Vol., No. 3, July 017 https://dx.doi.org/10.606/jaam.017.300 115 On the Local Convergence of Regula-falsi-type Method for Generalized Equations Farhana Alam

More information

Extreme Abridgment of Boyd and Vandenberghe s Convex Optimization

Extreme Abridgment of Boyd and Vandenberghe s Convex Optimization Extreme Abridgment of Boyd and Vandenberghe s Convex Optimization Compiled by David Rosenberg Abstract Boyd and Vandenberghe s Convex Optimization book is very well-written and a pleasure to read. The

More information

Partial regularity for fully nonlinear PDE

Partial regularity for fully nonlinear PDE Partial regularity for fully nonlinear PDE Luis Silvestre University of Chicago Joint work with Scott Armstrong and Charles Smart Outline Introduction Intro Review of fully nonlinear elliptic PDE Our result

More information

Sobolev regularity for the Monge-Ampère equation, with application to the semigeostrophic equations

Sobolev regularity for the Monge-Ampère equation, with application to the semigeostrophic equations Sobolev regularity for the Monge-Ampère equation, with application to the semigeostrophic equations Alessio Figalli Abstract In this note we review some recent results on the Sobolev regularity of solutions

More information

HESSIAN MEASURES III. Centre for Mathematics and Its Applications Australian National University Canberra, ACT 0200 Australia

HESSIAN MEASURES III. Centre for Mathematics and Its Applications Australian National University Canberra, ACT 0200 Australia HESSIAN MEASURES III Neil S. Trudinger Xu-Jia Wang Centre for Mathematics and Its Applications Australian National University Canberra, ACT 0200 Australia 1 HESSIAN MEASURES III Neil S. Trudinger Xu-Jia

More information

On the approximation of the principal eigenvalue for a class of nonlinear elliptic operators

On the approximation of the principal eigenvalue for a class of nonlinear elliptic operators On the approximation of the principal eigenvalue for a class of nonlinear elliptic operators Fabio Camilli ("Sapienza" Università di Roma) joint work with I.Birindelli ("Sapienza") I.Capuzzo Dolcetta ("Sapienza")

More information

g 2 (x) (1/3)M 1 = (1/3)(2/3)M.

g 2 (x) (1/3)M 1 = (1/3)(2/3)M. COMPACTNESS If C R n is closed and bounded, then by B-W it is sequentially compact: any sequence of points in C has a subsequence converging to a point in C Conversely, any sequentially compact C R n is

More information

Shiqian Ma, MAT-258A: Numerical Optimization 1. Chapter 4. Subgradient

Shiqian Ma, MAT-258A: Numerical Optimization 1. Chapter 4. Subgradient Shiqian Ma, MAT-258A: Numerical Optimization 1 Chapter 4 Subgradient Shiqian Ma, MAT-258A: Numerical Optimization 2 4.1. Subgradients definition subgradient calculus duality and optimality conditions Shiqian

More information

NONSMOOTH VARIANTS OF POWELL S BFGS CONVERGENCE THEOREM

NONSMOOTH VARIANTS OF POWELL S BFGS CONVERGENCE THEOREM NONSMOOTH VARIANTS OF POWELL S BFGS CONVERGENCE THEOREM JIAYI GUO AND A.S. LEWIS Abstract. The popular BFGS quasi-newton minimization algorithm under reasonable conditions converges globally on smooth

More information

Green s Functions and Distributions

Green s Functions and Distributions CHAPTER 9 Green s Functions and Distributions 9.1. Boundary Value Problems We would like to study, and solve if possible, boundary value problems such as the following: (1.1) u = f in U u = g on U, where

More information

Some Properties of the Augmented Lagrangian in Cone Constrained Optimization

Some Properties of the Augmented Lagrangian in Cone Constrained Optimization MATHEMATICS OF OPERATIONS RESEARCH Vol. 29, No. 3, August 2004, pp. 479 491 issn 0364-765X eissn 1526-5471 04 2903 0479 informs doi 10.1287/moor.1040.0103 2004 INFORMS Some Properties of the Augmented

More information

Chapter 2: Preliminaries and elements of convex analysis

Chapter 2: Preliminaries and elements of convex analysis Chapter 2: Preliminaries and elements of convex analysis Edoardo Amaldi DEIB Politecnico di Milano edoardo.amaldi@polimi.it Website: http://home.deib.polimi.it/amaldi/opt-14-15.shtml Academic year 2014-15

More information

Stationary isothermic surfaces and some characterizations of the hyperplane in the N-dimensional Euclidean space

Stationary isothermic surfaces and some characterizations of the hyperplane in the N-dimensional Euclidean space arxiv:081.165v1 [math.ap] 11 Dec 008 Stationary isothermic surfaces and some characterizations of the hyperplane in the N-dimensional Euclidean space Rolando Magnanini and Shigeru Sakaguchi October 6,

More information

SUPERCONVERGENCE PROPERTIES FOR OPTIMAL CONTROL PROBLEMS DISCRETIZED BY PIECEWISE LINEAR AND DISCONTINUOUS FUNCTIONS

SUPERCONVERGENCE PROPERTIES FOR OPTIMAL CONTROL PROBLEMS DISCRETIZED BY PIECEWISE LINEAR AND DISCONTINUOUS FUNCTIONS SUPERCONVERGENCE PROPERTIES FOR OPTIMAL CONTROL PROBLEMS DISCRETIZED BY PIECEWISE LINEAR AND DISCONTINUOUS FUNCTIONS A. RÖSCH AND R. SIMON Abstract. An optimal control problem for an elliptic equation

More information

McGill University Math 354: Honors Analysis 3

McGill University Math 354: Honors Analysis 3 Practice problems McGill University Math 354: Honors Analysis 3 not for credit Problem 1. Determine whether the family of F = {f n } functions f n (x) = x n is uniformly equicontinuous. 1st Solution: The

More information

Constrained Optimization and Lagrangian Duality

Constrained Optimization and Lagrangian Duality CIS 520: Machine Learning Oct 02, 2017 Constrained Optimization and Lagrangian Duality Lecturer: Shivani Agarwal Disclaimer: These notes are designed to be a supplement to the lecture. They may or may

More information

PIECEWISE LINEAR FINITE ELEMENT METHODS ARE NOT LOCALIZED

PIECEWISE LINEAR FINITE ELEMENT METHODS ARE NOT LOCALIZED PIECEWISE LINEAR FINITE ELEMENT METHODS ARE NOT LOCALIZED ALAN DEMLOW Abstract. Recent results of Schatz show that standard Galerkin finite element methods employing piecewise polynomial elements of degree

More information

Calculus of Variations

Calculus of Variations Calc. Var. (2004) DOI: 10.1007/s00526-004-0275-8 Calculus of Variations Ovidiu Savin The obstacle problem for Monge Ampere equation Received: 4 February 2003 / Accepted: 3 March 2004 Published online:

More information

Convexity of level sets for solutions to nonlinear elliptic problems in convex rings. Paola Cuoghi and Paolo Salani

Convexity of level sets for solutions to nonlinear elliptic problems in convex rings. Paola Cuoghi and Paolo Salani Convexity of level sets for solutions to nonlinear elliptic problems in convex rings Paola Cuoghi and Paolo Salani Dip.to di Matematica U. Dini - Firenze - Italy 1 Let u be a solution of a Dirichlet problem

More information

arxiv: v1 [math.na] 10 Apr 2008

arxiv: v1 [math.na] 10 Apr 2008 Approximating optimization problems over convex functions Néstor E. Aguilera Pedro Morin arxiv:0804.1693v1 [math.na] 10 Apr 2008 September 19, 2018 Abstract Many problems of theoretical and practical interest

More information

MATH 31BH Homework 1 Solutions

MATH 31BH Homework 1 Solutions MATH 3BH Homework Solutions January 0, 04 Problem.5. (a) (x, y)-plane in R 3 is closed and not open. To see that this plane is not open, notice that any ball around the origin (0, 0, 0) will contain points

More information

LECTURE 15: COMPLETENESS AND CONVEXITY

LECTURE 15: COMPLETENESS AND CONVEXITY LECTURE 15: COMPLETENESS AND CONVEXITY 1. The Hopf-Rinow Theorem Recall that a Riemannian manifold (M, g) is called geodesically complete if the maximal defining interval of any geodesic is R. On the other

More information

Division of the Humanities and Social Sciences. Supergradients. KC Border Fall 2001 v ::15.45

Division of the Humanities and Social Sciences. Supergradients. KC Border Fall 2001 v ::15.45 Division of the Humanities and Social Sciences Supergradients KC Border Fall 2001 1 The supergradient of a concave function There is a useful way to characterize the concavity of differentiable functions.

More information

Local strong convexity and local Lipschitz continuity of the gradient of convex functions

Local strong convexity and local Lipschitz continuity of the gradient of convex functions Local strong convexity and local Lipschitz continuity of the gradient of convex functions R. Goebel and R.T. Rockafellar May 23, 2007 Abstract. Given a pair of convex conjugate functions f and f, we investigate

More information

Regularity estimates for fully non linear elliptic equations which are asymptotically convex

Regularity estimates for fully non linear elliptic equations which are asymptotically convex Regularity estimates for fully non linear elliptic equations which are asymptotically convex Luis Silvestre and Eduardo V. Teixeira Abstract In this paper we deliver improved C 1,α regularity estimates

More information

MINIMAL GRAPHS PART I: EXISTENCE OF LIPSCHITZ WEAK SOLUTIONS TO THE DIRICHLET PROBLEM WITH C 2 BOUNDARY DATA

MINIMAL GRAPHS PART I: EXISTENCE OF LIPSCHITZ WEAK SOLUTIONS TO THE DIRICHLET PROBLEM WITH C 2 BOUNDARY DATA MINIMAL GRAPHS PART I: EXISTENCE OF LIPSCHITZ WEAK SOLUTIONS TO THE DIRICHLET PROBLEM WITH C 2 BOUNDARY DATA SPENCER HUGHES In these notes we prove that for any given smooth function on the boundary of

More information

Optimality Conditions for Nonsmooth Convex Optimization

Optimality Conditions for Nonsmooth Convex Optimization Optimality Conditions for Nonsmooth Convex Optimization Sangkyun Lee Oct 22, 2014 Let us consider a convex function f : R n R, where R is the extended real field, R := R {, + }, which is proper (f never

More information

1/12/05: sec 3.1 and my article: How good is the Lebesgue measure?, Math. Intelligencer 11(2) (1989),

1/12/05: sec 3.1 and my article: How good is the Lebesgue measure?, Math. Intelligencer 11(2) (1989), Real Analysis 2, Math 651, Spring 2005 April 26, 2005 1 Real Analysis 2, Math 651, Spring 2005 Krzysztof Chris Ciesielski 1/12/05: sec 3.1 and my article: How good is the Lebesgue measure?, Math. Intelligencer

More information

Subdifferential representation of convex functions: refinements and applications

Subdifferential representation of convex functions: refinements and applications Subdifferential representation of convex functions: refinements and applications Joël Benoist & Aris Daniilidis Abstract Every lower semicontinuous convex function can be represented through its subdifferential

More information

SYMMETRY IN REARRANGEMENT OPTIMIZATION PROBLEMS

SYMMETRY IN REARRANGEMENT OPTIMIZATION PROBLEMS Electronic Journal of Differential Equations, Vol. 2009(2009), No. 149, pp. 1 10. ISSN: 1072-6691. URL: http://ejde.math.txstate.edu or http://ejde.math.unt.edu ftp ejde.math.txstate.edu SYMMETRY IN REARRANGEMENT

More information

COMPARISON PRINCIPLES FOR CONSTRAINED SUBHARMONICS PH.D. COURSE - SPRING 2019 UNIVERSITÀ DI MILANO

COMPARISON PRINCIPLES FOR CONSTRAINED SUBHARMONICS PH.D. COURSE - SPRING 2019 UNIVERSITÀ DI MILANO COMPARISON PRINCIPLES FOR CONSTRAINED SUBHARMONICS PH.D. COURSE - SPRING 2019 UNIVERSITÀ DI MILANO KEVIN R. PAYNE 1. Introduction Constant coefficient differential inequalities and inclusions, constraint

More information

Math The Laplacian. 1 Green s Identities, Fundamental Solution

Math The Laplacian. 1 Green s Identities, Fundamental Solution Math. 209 The Laplacian Green s Identities, Fundamental Solution Let be a bounded open set in R n, n 2, with smooth boundary. The fact that the boundary is smooth means that at each point x the external

More information

minimize x subject to (x 2)(x 4) u,

minimize x subject to (x 2)(x 4) u, Math 6366/6367: Optimization and Variational Methods Sample Preliminary Exam Questions 1. Suppose that f : [, L] R is a C 2 -function with f () on (, L) and that you have explicit formulae for

More information

PROBLEMS. (b) (Polarization Identity) Show that in any inner product space

PROBLEMS. (b) (Polarization Identity) Show that in any inner product space 1 Professor Carl Cowen Math 54600 Fall 09 PROBLEMS 1. (Geometry in Inner Product Spaces) (a) (Parallelogram Law) Show that in any inner product space x + y 2 + x y 2 = 2( x 2 + y 2 ). (b) (Polarization

More information

FULL CHARACTERIZATION OF OPTIMAL TRANSPORT PLANS FOR CONCAVE COSTS

FULL CHARACTERIZATION OF OPTIMAL TRANSPORT PLANS FOR CONCAVE COSTS FULL CHARACTERIZATION OF OPTIMAL TRANSPORT PLANS FOR CONCAVE COSTS PAUL PEGON, DAVIDE PIAZZOLI, FILIPPO SANTAMBROGIO Abstract. This paper slightly improves a classical result by Gangbo and McCann (1996)

More information

Laplace s Equation. Chapter Mean Value Formulas

Laplace s Equation. Chapter Mean Value Formulas Chapter 1 Laplace s Equation Let be an open set in R n. A function u C 2 () is called harmonic in if it satisfies Laplace s equation n (1.1) u := D ii u = 0 in. i=1 A function u C 2 () is called subharmonic

More information

Numerical Analysis of Differential Equations Numerical Solution of Elliptic Boundary Value

Numerical Analysis of Differential Equations Numerical Solution of Elliptic Boundary Value Numerical Analysis of Differential Equations 188 5 Numerical Solution of Elliptic Boundary Value Problems 5 Numerical Solution of Elliptic Boundary Value Problems TU Bergakademie Freiberg, SS 2012 Numerical

More information

for all subintervals I J. If the same is true for the dyadic subintervals I D J only, we will write ϕ BMO d (J). In fact, the following is true

for all subintervals I J. If the same is true for the dyadic subintervals I D J only, we will write ϕ BMO d (J). In fact, the following is true 3 ohn Nirenberg inequality, Part I A function ϕ L () belongs to the space BMO() if sup ϕ(s) ϕ I I I < for all subintervals I If the same is true for the dyadic subintervals I D only, we will write ϕ BMO

More information

Asymptotic Behavior of Infinity Harmonic Functions Near an Isolated Singularity

Asymptotic Behavior of Infinity Harmonic Functions Near an Isolated Singularity Savin, O., and C. Wang. (2008) Asymptotic Behavior of Infinity Harmonic Functions, International Mathematics Research Notices, Vol. 2008, Article ID rnm163, 23 pages. doi:10.1093/imrn/rnm163 Asymptotic

More information

Convex Optimization Theory. Chapter 5 Exercises and Solutions: Extended Version

Convex Optimization Theory. Chapter 5 Exercises and Solutions: Extended Version Convex Optimization Theory Chapter 5 Exercises and Solutions: Extended Version Dimitri P. Bertsekas Massachusetts Institute of Technology Athena Scientific, Belmont, Massachusetts http://www.athenasc.com

More information

C 1 regularity of solutions of the Monge-Ampère equation for optimal transport in dimension two

C 1 regularity of solutions of the Monge-Ampère equation for optimal transport in dimension two C 1 regularity of solutions of the Monge-Ampère equation for optimal transport in dimension two Alessio Figalli, Grégoire Loeper Abstract We prove C 1 regularity of c-convex weak Alexandrov solutions of

More information

Real Analysis Math 131AH Rudin, Chapter #1. Dominique Abdi

Real Analysis Math 131AH Rudin, Chapter #1. Dominique Abdi Real Analysis Math 3AH Rudin, Chapter # Dominique Abdi.. If r is rational (r 0) and x is irrational, prove that r + x and rx are irrational. Solution. Assume the contrary, that r+x and rx are rational.

More information

Lecture Notes: African Institute of Mathematics Senegal, January Topic Title: A short introduction to numerical methods for elliptic PDEs

Lecture Notes: African Institute of Mathematics Senegal, January Topic Title: A short introduction to numerical methods for elliptic PDEs Lecture Notes: African Institute of Mathematics Senegal, January 26 opic itle: A short introduction to numerical methods for elliptic PDEs Authors and Lecturers: Gerard Awanou (University of Illinois-Chicago)

More information

Assignment 1: From the Definition of Convexity to Helley Theorem

Assignment 1: From the Definition of Convexity to Helley Theorem Assignment 1: From the Definition of Convexity to Helley Theorem Exercise 1 Mark in the following list the sets which are convex: 1. {x R 2 : x 1 + i 2 x 2 1, i = 1,..., 10} 2. {x R 2 : x 2 1 + 2ix 1x

More information

i=1 α i. Given an m-times continuously

i=1 α i. Given an m-times continuously 1 Fundamentals 1.1 Classification and characteristics Let Ω R d, d N, d 2, be an open set and α = (α 1,, α d ) T N d 0, N 0 := N {0}, a multiindex with α := d i=1 α i. Given an m-times continuously differentiable

More information

LORENTZ ESTIMATES FOR ASYMPTOTICALLY REGULAR FULLY NONLINEAR ELLIPTIC EQUATIONS

LORENTZ ESTIMATES FOR ASYMPTOTICALLY REGULAR FULLY NONLINEAR ELLIPTIC EQUATIONS Electronic Journal of Differential Equations, Vol. 27 27), No. 2, pp. 3. ISSN: 72-669. URL: http://ejde.math.txstate.edu or http://ejde.math.unt.edu LORENTZ ESTIMATES FOR ASYMPTOTICALLY REGULAR FULLY NONLINEAR

More information

ASYMPTOTICALLY EXACT A POSTERIORI ESTIMATORS FOR THE POINTWISE GRADIENT ERROR ON EACH ELEMENT IN IRREGULAR MESHES. PART II: THE PIECEWISE LINEAR CASE

ASYMPTOTICALLY EXACT A POSTERIORI ESTIMATORS FOR THE POINTWISE GRADIENT ERROR ON EACH ELEMENT IN IRREGULAR MESHES. PART II: THE PIECEWISE LINEAR CASE MATEMATICS OF COMPUTATION Volume 73, Number 246, Pages 517 523 S 0025-5718(0301570-9 Article electronically published on June 17, 2003 ASYMPTOTICALLY EXACT A POSTERIORI ESTIMATORS FOR TE POINTWISE GRADIENT

More information

Chapter 2 Convex Analysis

Chapter 2 Convex Analysis Chapter 2 Convex Analysis The theory of nonsmooth analysis is based on convex analysis. Thus, we start this chapter by giving basic concepts and results of convexity (for further readings see also [202,

More information

On self-concordant barriers for generalized power cones

On self-concordant barriers for generalized power cones On self-concordant barriers for generalized power cones Scott Roy Lin Xiao January 30, 2018 Abstract In the study of interior-point methods for nonsymmetric conic optimization and their applications, Nesterov

More information

GEOMETRIC APPROACH TO CONVEX SUBDIFFERENTIAL CALCULUS October 10, Dedicated to Franco Giannessi and Diethard Pallaschke with great respect

GEOMETRIC APPROACH TO CONVEX SUBDIFFERENTIAL CALCULUS October 10, Dedicated to Franco Giannessi and Diethard Pallaschke with great respect GEOMETRIC APPROACH TO CONVEX SUBDIFFERENTIAL CALCULUS October 10, 2018 BORIS S. MORDUKHOVICH 1 and NGUYEN MAU NAM 2 Dedicated to Franco Giannessi and Diethard Pallaschke with great respect Abstract. In

More information

MATH 202B - Problem Set 5

MATH 202B - Problem Set 5 MATH 202B - Problem Set 5 Walid Krichene (23265217) March 6, 2013 (5.1) Show that there exists a continuous function F : [0, 1] R which is monotonic on no interval of positive length. proof We know there

More information

ON SOME ELLIPTIC PROBLEMS IN UNBOUNDED DOMAINS

ON SOME ELLIPTIC PROBLEMS IN UNBOUNDED DOMAINS Chin. Ann. Math.??B(?), 200?, 1 20 DOI: 10.1007/s11401-007-0001-x ON SOME ELLIPTIC PROBLEMS IN UNBOUNDED DOMAINS Michel CHIPOT Abstract We present a method allowing to obtain existence of a solution for

More information

Inverse problems Total Variation Regularization Mark van Kraaij Casa seminar 23 May 2007 Technische Universiteit Eindh ove n University of Technology

Inverse problems Total Variation Regularization Mark van Kraaij Casa seminar 23 May 2007 Technische Universiteit Eindh ove n University of Technology Inverse problems Total Variation Regularization Mark van Kraaij Casa seminar 23 May 27 Introduction Fredholm first kind integral equation of convolution type in one space dimension: g(x) = 1 k(x x )f(x

More information

Test 2 Review Math 1111 College Algebra

Test 2 Review Math 1111 College Algebra Test 2 Review Math 1111 College Algebra 1. Begin by graphing the standard quadratic function f(x) = x 2. Then use transformations of this graph to graph the given function. g(x) = x 2 + 2 *a. b. c. d.

More information

The Lusin Theorem and Horizontal Graphs in the Heisenberg Group

The Lusin Theorem and Horizontal Graphs in the Heisenberg Group Analysis and Geometry in Metric Spaces Research Article DOI: 10.2478/agms-2013-0008 AGMS 2013 295-301 The Lusin Theorem and Horizontal Graphs in the Heisenberg Group Abstract In this paper we prove that

More information

Mathematics II, course

Mathematics II, course Mathematics II, course 2013-2014 Juan Pablo Rincón Zapatero October 24, 2013 Summary: The course has four parts that we describe below. (I) Topology in Rn is a brief review of the main concepts and properties

More information

Applied Mathematics Letters

Applied Mathematics Letters Applied Mathematics Letters 25 (2012) 974 979 Contents lists available at SciVerse ScienceDirect Applied Mathematics Letters journal homepage: www.elsevier.com/locate/aml On dual vector equilibrium problems

More information

MEAN CURVATURE FLOW OF ENTIRE GRAPHS EVOLVING AWAY FROM THE HEAT FLOW

MEAN CURVATURE FLOW OF ENTIRE GRAPHS EVOLVING AWAY FROM THE HEAT FLOW MEAN CURVATURE FLOW OF ENTIRE GRAPHS EVOLVING AWAY FROM THE HEAT FLOW GREGORY DRUGAN AND XUAN HIEN NGUYEN Abstract. We present two initial graphs over the entire R n, n 2 for which the mean curvature flow

More information

A PREDICTOR-CORRECTOR PATH-FOLLOWING ALGORITHM FOR SYMMETRIC OPTIMIZATION BASED ON DARVAY'S TECHNIQUE

A PREDICTOR-CORRECTOR PATH-FOLLOWING ALGORITHM FOR SYMMETRIC OPTIMIZATION BASED ON DARVAY'S TECHNIQUE Yugoslav Journal of Operations Research 24 (2014) Number 1, 35-51 DOI: 10.2298/YJOR120904016K A PREDICTOR-CORRECTOR PATH-FOLLOWING ALGORITHM FOR SYMMETRIC OPTIMIZATION BASED ON DARVAY'S TECHNIQUE BEHROUZ

More information

Geometric problems. Chapter Projection on a set. The distance of a point x 0 R n to a closed set C R n, in the norm, is defined as

Geometric problems. Chapter Projection on a set. The distance of a point x 0 R n to a closed set C R n, in the norm, is defined as Chapter 8 Geometric problems 8.1 Projection on a set The distance of a point x 0 R n to a closed set C R n, in the norm, is defined as dist(x 0,C) = inf{ x 0 x x C}. The infimum here is always achieved.

More information

Integro-differential equations: Regularity theory and Pohozaev identities

Integro-differential equations: Regularity theory and Pohozaev identities Integro-differential equations: Regularity theory and Pohozaev identities Xavier Ros Oton Departament Matemàtica Aplicada I, Universitat Politècnica de Catalunya PhD Thesis Advisor: Xavier Cabré Xavier

More information

EXISTENCE OF SOLUTIONS FOR KIRCHHOFF TYPE EQUATIONS WITH UNBOUNDED POTENTIAL. 1. Introduction In this article, we consider the Kirchhoff type problem

EXISTENCE OF SOLUTIONS FOR KIRCHHOFF TYPE EQUATIONS WITH UNBOUNDED POTENTIAL. 1. Introduction In this article, we consider the Kirchhoff type problem Electronic Journal of Differential Equations, Vol. 207 (207), No. 84, pp. 2. ISSN: 072-669. URL: http://ejde.math.txstate.edu or http://ejde.math.unt.edu EXISTENCE OF SOLUTIONS FOR KIRCHHOFF TYPE EQUATIONS

More information

The Relation Between Pseudonormality and Quasiregularity in Constrained Optimization 1

The Relation Between Pseudonormality and Quasiregularity in Constrained Optimization 1 October 2003 The Relation Between Pseudonormality and Quasiregularity in Constrained Optimization 1 by Asuman E. Ozdaglar and Dimitri P. Bertsekas 2 Abstract We consider optimization problems with equality,

More information

SELF-ADJOINTNESS OF SCHRÖDINGER-TYPE OPERATORS WITH SINGULAR POTENTIALS ON MANIFOLDS OF BOUNDED GEOMETRY

SELF-ADJOINTNESS OF SCHRÖDINGER-TYPE OPERATORS WITH SINGULAR POTENTIALS ON MANIFOLDS OF BOUNDED GEOMETRY Electronic Journal of Differential Equations, Vol. 2003(2003), No.??, pp. 1 8. ISSN: 1072-6691. URL: http://ejde.math.swt.edu or http://ejde.math.unt.edu ftp ejde.math.swt.edu (login: ftp) SELF-ADJOINTNESS

More information

Convex Functions and Optimization

Convex Functions and Optimization Chapter 5 Convex Functions and Optimization 5.1 Convex Functions Our next topic is that of convex functions. Again, we will concentrate on the context of a map f : R n R although the situation can be generalized

More information

VISCOSITY SOLUTIONS OF ELLIPTIC EQUATIONS

VISCOSITY SOLUTIONS OF ELLIPTIC EQUATIONS VISCOSITY SOLUTIONS OF ELLIPTIC EQUATIONS LUIS SILVESTRE These are the notes from the summer course given in the Second Chicago Summer School In Analysis, in June 2015. We introduce the notion of viscosity

More information

THE REGULARITY OF MAPPINGS WITH A CONVEX POTENTIAL LUIS A. CAFFARELLI

THE REGULARITY OF MAPPINGS WITH A CONVEX POTENTIAL LUIS A. CAFFARELLI JOURNAL OF THE AMERICAN MATHEMATICAL SOCIETY Volume 5, Number I, Jaouary 1992 THE REGULARITY OF MAPPINGS WITH A CONVEX POTENTIAL LUIS A. CAFFARELLI In this work, we apply the techniques developed in [Cl]

More information

REGULARITY OF POTENTIAL FUNCTIONS IN OPTIMAL TRANSPORTATION. Centre for Mathematics and Its Applications The Australian National University

REGULARITY OF POTENTIAL FUNCTIONS IN OPTIMAL TRANSPORTATION. Centre for Mathematics and Its Applications The Australian National University ON STRICT CONVEXITY AND C 1 REGULARITY OF POTENTIAL FUNCTIONS IN OPTIMAL TRANSPORTATION Neil Trudinger Xu-Jia Wang Centre for Mathematics and Its Applications The Australian National University Abstract.

More information

Functional Analysis. Franck Sueur Metric spaces Definitions Completeness Compactness Separability...

Functional Analysis. Franck Sueur Metric spaces Definitions Completeness Compactness Separability... Functional Analysis Franck Sueur 2018-2019 Contents 1 Metric spaces 1 1.1 Definitions........................................ 1 1.2 Completeness...................................... 3 1.3 Compactness......................................

More information

LARGE DEVIATIONS OF TYPICAL LINEAR FUNCTIONALS ON A CONVEX BODY WITH UNCONDITIONAL BASIS. S. G. Bobkov and F. L. Nazarov. September 25, 2011

LARGE DEVIATIONS OF TYPICAL LINEAR FUNCTIONALS ON A CONVEX BODY WITH UNCONDITIONAL BASIS. S. G. Bobkov and F. L. Nazarov. September 25, 2011 LARGE DEVIATIONS OF TYPICAL LINEAR FUNCTIONALS ON A CONVEX BODY WITH UNCONDITIONAL BASIS S. G. Bobkov and F. L. Nazarov September 25, 20 Abstract We study large deviations of linear functionals on an isotropic

More information

Optimization and Optimal Control in Banach Spaces

Optimization and Optimal Control in Banach Spaces Optimization and Optimal Control in Banach Spaces Bernhard Schmitzer October 19, 2017 1 Convex non-smooth optimization with proximal operators Remark 1.1 (Motivation). Convex optimization: easier to solve,

More information

VISCOSITY SOLUTIONS. We follow Han and Lin, Elliptic Partial Differential Equations, 5.

VISCOSITY SOLUTIONS. We follow Han and Lin, Elliptic Partial Differential Equations, 5. VISCOSITY SOLUTIONS PETER HINTZ We follow Han and Lin, Elliptic Partial Differential Equations, 5. 1. Motivation Throughout, we will assume that Ω R n is a bounded and connected domain and that a ij C(Ω)

More information

Notions such as convergent sequence and Cauchy sequence make sense for any metric space. Convergent Sequences are Cauchy

Notions such as convergent sequence and Cauchy sequence make sense for any metric space. Convergent Sequences are Cauchy Banach Spaces These notes provide an introduction to Banach spaces, which are complete normed vector spaces. For the purposes of these notes, all vector spaces are assumed to be over the real numbers.

More information

Second Order Elliptic PDE

Second Order Elliptic PDE Second Order Elliptic PDE T. Muthukumar tmk@iitk.ac.in December 16, 2014 Contents 1 A Quick Introduction to PDE 1 2 Classification of Second Order PDE 3 3 Linear Second Order Elliptic Operators 4 4 Periodic

More information

ON THE REGULARITY OF SAMPLE PATHS OF SUB-ELLIPTIC DIFFUSIONS ON MANIFOLDS

ON THE REGULARITY OF SAMPLE PATHS OF SUB-ELLIPTIC DIFFUSIONS ON MANIFOLDS Bendikov, A. and Saloff-Coste, L. Osaka J. Math. 4 (5), 677 7 ON THE REGULARITY OF SAMPLE PATHS OF SUB-ELLIPTIC DIFFUSIONS ON MANIFOLDS ALEXANDER BENDIKOV and LAURENT SALOFF-COSTE (Received March 4, 4)

More information

Mean Field Games on networks

Mean Field Games on networks Mean Field Games on networks Claudio Marchi Università di Padova joint works with: S. Cacace (Rome) and F. Camilli (Rome) C. Marchi (Univ. of Padova) Mean Field Games on networks Roma, June 14 th, 2017

More information

Kaisa Joki Adil M. Bagirov Napsu Karmitsa Marko M. Mäkelä. New Proximal Bundle Method for Nonsmooth DC Optimization

Kaisa Joki Adil M. Bagirov Napsu Karmitsa Marko M. Mäkelä. New Proximal Bundle Method for Nonsmooth DC Optimization Kaisa Joki Adil M. Bagirov Napsu Karmitsa Marko M. Mäkelä New Proximal Bundle Method for Nonsmooth DC Optimization TUCS Technical Report No 1130, February 2015 New Proximal Bundle Method for Nonsmooth

More information

W 2,1 REGULARITY FOR SOLUTIONS OF THE MONGE-AMPÈRE EQUATION

W 2,1 REGULARITY FOR SOLUTIONS OF THE MONGE-AMPÈRE EQUATION W 2,1 REGULARITY FOR SOLUTIONS OF THE MONGE-AMPÈRE EQUATION GUIDO DE PHILIPPIS AND ALESSIO FIGALLI Abstract. In this paper we prove that a strictly convex Alexandrov solution u of the Monge-Ampère equation,

More information

Math General Topology Fall 2012 Homework 6 Solutions

Math General Topology Fall 2012 Homework 6 Solutions Math 535 - General Topology Fall 202 Homework 6 Solutions Problem. Let F be the field R or C of real or complex numbers. Let n and denote by F[x, x 2,..., x n ] the set of all polynomials in n variables

More information

10-725/ Optimization Midterm Exam

10-725/ Optimization Midterm Exam 10-725/36-725 Optimization Midterm Exam November 6, 2012 NAME: ANDREW ID: Instructions: This exam is 1hr 20mins long Except for a single two-sided sheet of notes, no other material or discussion is permitted

More information

Math 421, Homework #9 Solutions

Math 421, Homework #9 Solutions Math 41, Homework #9 Solutions (1) (a) A set E R n is said to be path connected if for any pair of points x E and y E there exists a continuous function γ : [0, 1] R n satisfying γ(0) = x, γ(1) = y, and

More information

Introduction to Real Analysis Alternative Chapter 1

Introduction to Real Analysis Alternative Chapter 1 Christopher Heil Introduction to Real Analysis Alternative Chapter 1 A Primer on Norms and Banach Spaces Last Updated: March 10, 2018 c 2018 by Christopher Heil Chapter 1 A Primer on Norms and Banach Spaces

More information

ON GENERALIZED-CONVEX CONSTRAINED MULTI-OBJECTIVE OPTIMIZATION

ON GENERALIZED-CONVEX CONSTRAINED MULTI-OBJECTIVE OPTIMIZATION ON GENERALIZED-CONVEX CONSTRAINED MULTI-OBJECTIVE OPTIMIZATION CHRISTIAN GÜNTHER AND CHRISTIANE TAMMER Abstract. In this paper, we consider multi-objective optimization problems involving not necessarily

More information

On intermediate value theorem in ordered Banach spaces for noncompact and discontinuous mappings

On intermediate value theorem in ordered Banach spaces for noncompact and discontinuous mappings Int. J. Nonlinear Anal. Appl. 7 (2016) No. 1, 295-300 ISSN: 2008-6822 (electronic) http://dx.doi.org/10.22075/ijnaa.2015.341 On intermediate value theorem in ordered Banach spaces for noncompact and discontinuous

More information

The optimal partial transport problem

The optimal partial transport problem The optimal partial transport problem Alessio Figalli Abstract Given two densities f and g, we consider the problem of transporting a fraction m [0, min{ f L 1, g L 1}] of the mass of f onto g minimizing

More information

Measure and Integration: Solutions of CW2

Measure and Integration: Solutions of CW2 Measure and Integration: s of CW2 Fall 206 [G. Holzegel] December 9, 206 Problem of Sheet 5 a) Left (f n ) and (g n ) be sequences of integrable functions with f n (x) f (x) and g n (x) g (x) for almost

More information