An optimal transportation problem with import/export taxes on the boundary


 Reynold Barrett
 1 years ago
 Views:
Transcription
1 An optimal transportation problem with import/export taxes on the boundary Julián Toledo Workshop International sur les Mathématiques et l Environnement Essaouira, November Joint work with J. M. Mazón and J. D. Rossi
2 Schedule 1. Statement of the problem 2. Mass Transport Theory 3. The mass transport problem with import/export taxes: Duality 4. EvansGangbo Performance 5. Optimal Transport Plans 6. Limited importation/exportation
3 Statement of the problem
4 Business A business man (BM) produces some product in some factories in : f + gives the amount of product and its location. There are some consumers of the product in : f gives its distribution. BM wants: to transport all the mass f + to f (to satisfy the consumers) or to the boundary (export).
5 Costs and constrains The BM pays: the transport costs (given by the Euclidean distance) PLUS when a unit of mass is left on a point y : T e (y), the export taxes. BM must satisfy all the demand of the consumers. So that, he has to import product, if necessary, from the exterior, paying: T i (x), import taxes, for each unit of mass that enters from x.
6 Some freedom BM has the freedom to chose to export or import mass provided he transports all the mass in f + and covers all the mass of f.
7 MAIN GOAL The main goal here is to minimize the total cost of this operation, that is given by the transport cost PLUS export/import taxes. Given the set of transport plans { A(f +, f ) := µ M + ( ) : π 1 #µ = f + L N, π 2 #µ = f L N }, the goal is to obtain min µ A(f +,f ) x y dµ + T i d(π 1 #µ) + T e d(π 2 #µ) π i : R N R N are the projections: π 1 (x, y) := x, π 2 (x, y) := y. Given a Radon measure γ in X X, π 1 #γ is the marginal proj x (γ), and π 2 #γ is the marginal proj y (γ).
8 Equilibrium of masses The transport must verify the natural equilibrium of masses, nevertheless it is not necessary to impose f + (x) dx = f (y) dy.
9 Equilibrium of masses The transport must verify the natural equilibrium of masses, nevertheless it is not necessary to impose f + (x) dx = f (y) dy. In this transport problem two masses appear on, that are unknowns, the ones that encode the mass exported and the mass imported. Taking into account this masses we must have the equilibrium condition.
10 Conditions on the domains is assumed to be convex: we prevent the fact that the boundary of is crossed when the mass is transported inside.
11 Conditions on the domains is assumed to be convex: we prevent the fact that the boundary of is crossed when the mass is transported inside. On the support of the f ± we assume that supp(f + ) supp(f ) =.
12 Mass Transport Theory
13 Monge problem Given two measures µ, ν M(X ) (say X R N ) satisfying the mass balance condition dµ = dν, Monge s Problem consists in solving x T (x) dµ(x), inf T #µ=ν X X T : X R N are Borel functions pushing forward µ to ν: T #µ = ν, T #µ(b) = µ(t 1 (B)). X A minimizer T is called an optimal transport map of µ to ν. In the case that µ and ν represent the distribution for production and consumption of some commodity, the problem is to decide which producer should supply each consumer minimizing the total transport cost.
14 Relaxed Problem: MongeKantorovich Problem In general, the Monge problem is illposed. But...
15 Relaxed Problem: MongeKantorovich Problem In general, the Monge problem is illposed. But... Fix two measures µ, ν satisfying the mass balance condition. Let Π(µ, ν) the set of transport plans γ between µ and ν: proj x (γ) = µ, proj y (γ) = ν. The problem of finding a measure γ Π(µ, ν) which minimizes the cost functional K(γ) := x y dγ(x, y), in the set Π(µ, ν), is wellposed. X X A minimizer γ is called an optimal transport plan between µ and ν. Linearity makes the MongeKantorovich problem simpler than Monge original problem: a continuitycompactness argument guarantees the existence of an optimal transport plan.
16 Dual Problem: Kantorovich Duality Linear minimization problems admit a dual formulation: For (ϕ, ψ) L 1 (dµ) L 1 (dν), define J(ϕ, ψ) := ϕ dµ + ψ dν, X X and let Φ be the set of all measurable functions (ϕ, ψ) L 1 (dµ) L 1 (dν) satisfying Then ϕ(x) + ψ(y) x y for µ ν almost all (x, y) X X. inf K(γ) = sup J(ϕ, ψ). γ Π(µ,ν) (ϕ,ψ) Φ
17 And... KantorovichRubinstein Theorem Theorem (KantorovichRubinstein) where min K(γ) = γ Π(µ,ν) sup u K 1(X ) X u d(µ ν), K 1 (X ) := {u : X R : u(x) u(y) x y x, y X } is the set of 1Lipschitz functions in X. The maximizers u of the right hand side are called Kantorovich (transport) potentials.
18 Evans and Gangbo Approach For µ = f + L N and ν = f L N, f + and f adequate Lebesgue integrable functions, Evans and Gangbo find a Kantorovich potential as a limit, as p, of solutions to { p u p = f + f in B(0, R), u p = 0 on B(0, R). Theorem u p u K 1 () as p. u (x)(f + (x) f (x)) dx = u(x)(f + (x) f (x)) dx. max u K 1()
19 Evans and Gangbo Approach For µ = f + L N and ν = f L N, f + and f adequate Lebesgue integrable functions, Evans and Gangbo find a Kantorovich potential as a limit, as p, of solutions to { p u p = f + f in B(0, R), u p = 0 on B(0, R). Theorem u p u K 1 () as p. u (x)(f + (x) f (x)) dx = u(x)(f + (x) f (x)) dx. max u K 1() Characterization of the Kantorovich potential: There exists 0 a L () such that f + f = div(a u ) in D (). Furthermore u = 1 a.e. in the set {a > 0}.
20 Evans and Gangbo Approach Evans and Gangbo used this PDE to find a proof of the existence of an optimal transport map for the classical Monge problem, different to the first one given by Sudakov in 1979 by means of probability methods.
21 The mass transport problem with import/export taxes First approach: Duality
22 A Clever Fellow
23 A Clever Fellow A clever fellow proposes the BM to leave him the planing and offers him the following deal:
24 A Clever Fellow A clever fellow proposes the BM to leave him the planing and offers him the following deal: * to pick up the product f + (x), x, he will charge him ϕ(x), * to pick it up at x, T i (x) (paying the taxes by himself), * to leave the product at the consumer s location f (y), y, he will charge him ψ(y), and * for leaving it at y, T e (y) (paying again the taxes himself).
25 A Clever Fellow A clever fellow proposes the BM to leave him the planing and offers him the following deal: * to pick up the product f + (x), x, he will charge him ϕ(x), * to pick it up at x, T i (x) (paying the taxes by himself), * to leave the product at the consumer s location f (y), y, he will charge him ψ(y), and * for leaving it at y, T e (y) (paying again the taxes himself). * He will do all that job in such a way that: ϕ(x) + ψ(y) x y T i ϕ and T e ψ on. (assuming negative payments if necessary).
26 A Clever Fellow A clever fellow proposes the BM to leave him the planing and offers him the following deal: * to pick up the product f + (x), x, he will charge him ϕ(x), * to pick it up at x, T i (x) (paying the taxes by himself), * to leave the product at the consumer s location f (y), y, he will charge him ψ(y), and * for leaving it at y, T e (y) (paying again the taxes himself). * He will do all that job in such a way that: ϕ(x) + ψ(y) x y T i ϕ and T e ψ on. (assuming negative payments if necessary). Observe that we need to impose the following condition: T i (x) T e (y) x y, x, y, (there is not benefit of importing mass from x and exporting it to y). For x = y : T i (x) + T e (x) 0 means that in the same point the sum of exportation and importation taxes is non negative.
27 Clever Fellow s Goal Consider J(ϕ, ψ) := ϕ(x)f + (x) dx + ψ(y)f (y) dy, defined for continuous functions, and let { (ϕ, ψ) C() C() : ϕ(x) + ψ(y) x y, B(T i, T e ) := T i ϕ, T e ψ on The aim of the fellow that helps the business man is to obtain sup {J(ϕ, ψ) : (ϕ, ψ) B(T i, T e )}. }.
28 Clever Fellow s Goal Consider J(ϕ, ψ) := ϕ(x)f + (x) dx + ψ(y)f (y) dy, defined for continuous functions, and let { (ϕ, ψ) C() C() : ϕ(x) + ψ(y) x y, B(T i, T e ) := T i ϕ, T e ψ on The aim of the fellow that helps the business man is to obtain sup {J(ϕ, ψ) : (ϕ, ψ) B(T i, T e )}. Now, given (ϕ, ψ) B(T i, T e ) and µ A(f +, f ): J(ϕ, ψ) = ϕ(x)f + (x) dx + ψ(y)f (y) dy = ϕ(x)dπ 1 #µ ϕdπ 1 #µ + ψ(y)dπ 2 #µ ψdπ 2 #µ }.
29 Clever Fellow s Goal Consider J(ϕ, ψ) := ϕ(x)f + (x) dx + ψ(y)f (y) dy, defined for continuous functions, and let { (ϕ, ψ) C() C() : ϕ(x) + ψ(y) x y, B(T i, T e ) := T i ϕ, T e ψ on The aim of the fellow that helps the business man is to obtain sup {J(ϕ, ψ) : (ϕ, ψ) B(T i, T e )}. Now, given (ϕ, ψ) B(T i, T e ) and µ A(f +, f ): J(ϕ, ψ) = ϕ(x)f + (x) dx + ψ(y)f (y) dy = ϕ(x)dπ 1 #µ ϕdπ 1 #µ + ψ(y)dπ 2 #µ ψdπ 2 #µ x y dµ + T i d(π 1 #µ) + T e d(π 2 #µ). }.
30 Acceptance of the deal Therefore, sup (ϕ,ψ) B(T i,t e) inf µ A(f +,f ) J(ϕ, ψ) x y dµ + T i d(π 1 #µ) + T e d(π 2 #µ).
31 Acceptance of the deal Therefore, sup (ϕ,ψ) B(T i,t e) inf µ A(f +,f ) J(ϕ, ψ) x y dµ + T i d(π 1 #µ) + T e d(π 2 #µ). This inequality will imply that the business man accept the offer.
32 NO GAP But, in fact, there is no gap between both costs: Theorem Assume that T i and T e satisfy T i (x) T e (y) < x y x, y. Then, sup J(ϕ, ψ) (ϕ,ψ) B(T i,t e) = min µ A(f +,f ) x y dµ + T i dπ 1 #µ + T e dπ 2 #µ. Proof based on FenchelRocafellar s duality Theorem.
33 UNKNOWNS There is an important difference between the problem we are studying and the classical transport problem: there are masses, the ones that appear on the boundary, that are unknown variables.
34 UNKNOWNS There is an important difference between the problem we are studying and the classical transport problem: there are masses, the ones that appear on the boundary, that are unknown variables. The above result is an abstract result and does not give explicitly the measures involved in the problem. But...
35 EvansGangbo Performance Kantorovich potentials and optimal export/import masses
36 Let f L () and N < p < +. Let g i C( ), g 1 g 2 on. Set Wg 1,p 1,g 2 () = {u W 1,p () : g 1 u g 2 on } and consider the energy functional u(x) p Ψ p (u) := dx f (x)u(x) dx. p
37 Let f L () and N < p < +. Let g i C( ), g 1 g 2 on. Set Wg 1,p 1,g 2 () = {u W 1,p () : g 1 u g 2 on } and consider the energy functional u(x) p Ψ p (u) := dx f (x)u(x) dx. p Wg 1,p 1,g 2 () is a closed convex subset of W 1,p () and Ψ p is convex, l.s.c. and coercive. Then min Ψ p (u) u Wg 1,p 1,g 2 () has a minimizer u p in W 1,p g 1,g 2 (), which is a least energy solution of the obstacle problem { p u = f in, g 1 u g 2 on.
38 Theorem Assume g 1, g 2 satisfy g 1 (x) g 2 (y) x y x, y. Then, up to subsequence, u p u uniformly, and u is a maximizer of the variational problem { } max w(x)f (x) dx : w Wg 1, 1,g 2 (), w L () 1.
39 Theorem Assume g 1, g 2 satisfy g 1 (x) g 2 (y) x y x, y. Then, up to subsequence, u p u uniformly, and u is a maximizer of the variational problem { } max w(x)f (x) dx : w Wg 1, 1,g 2 (), w L () 1. A natural question: Let f + and f be the positive and negative part of f. Can u be interpreted as a kind of Kantorovich potential for some transport problem involving f + and f?
40 DUALITY Theorem If then g 1 (x) g 2 (x) x y x, y, u (x)(f + (x) f (x))dx = sup J(ϕ, ψ) (ϕ,ψ) B( g 1,g 2) = min µ A(f +,f ) x y dµ g 1 dπ 1 #µ + where u is the maximizer given in the above Theorem. g 2 dπ 2 #µ,
41 DUALITY Theorem If then g 1 (x) g 2 (x) x y x, y, u (x)(f + (x) f (x))dx = sup J(ϕ, ψ) (ϕ,ψ) B( g 1,g 2) = min µ A(f +,f ) x y dµ g 1 dπ 1 #µ + where u is the maximizer given in the above Theorem. g 2 dπ 2 #µ, Observe that setting T i = g 1 and T e = g 2 we are dealing with our transport problem.
42 Key result Theorem Assume g 1 (x) g 2 (y) < x y x, y. Let u p, u the functions obtained above. Then, up to a subsequence, X p := Du p p 2 Du p X weakly* in the sense of measures, div(x ) = f in the sense of distributions in, Moreover, the distributions X p η defined as X p η, ϕ := X p ϕ f ϕ for ϕ C0 (R N ), are Radon measures supported on, that, up to a subsequence, X p η V weakly* in the sense of measures; and...
43 Key result u is a Kantorovich potential for the classical transport problem of the measure ˆf + = f + L N + V + to ˆf = f L N + V. Moreover, u = g 1 on supp(v + ), u = g 2 on supp(v ).
44 Key result This result provides a proof for the previous duality Theorem: u d(ˆf + ˆf ) = min x y dν = x y dν 0, ν Π(ˆf +,ˆf ) for some ν 0 Π(ˆf +, ˆf ). Then, since π 1 #ν 0 = V + and π 2 #ν 0 = V, u (f + f ) = x y dν 0 g 1 dπ 1 #ν 0 + g 2 dπ 2 #ν 0.
45 Moreover, we have that V + and V are import and export masses in our original problem.
46 Proof Let u p be a minimizer of Ψ p (u) := u(x) p p dx f (x)u(x) dx. Then, { p u p = f in, g 1 u p g 2 on.
47 Proof Let u p be a minimizer of Ψ p (u) := u(x) p p dx f (x)u(x) dx. Then, { p u p = f in, g 1 u p g 2 on. Therefore for X p := Du p p 2 Du p we have that div(x p ) = f in the sense of distributions, and that the distribution X p η, defined as X p η, ϕ := X p ϕ f ϕ for ϕ C0 (R N ), is supported on.
48 We proof that, in fact, X p η is the sum of a nonnegative and a nonpositive distribution, so it is a measure, and moreover that supp((x p η) + ) {x : u p (x) = g 1 (x)}, and supp((x p η) ) {x : u p (x) = g 2 (x)}. In addition, we have that X p ϕ = f ϕ + ϕ d (X p η) ϕ W 1,p ().
49 Boundedness of u p, X p and X p η By using adequate test functions we get that, for p N + 1, u p p C, C constant not depending on p. d(x p η) ± C,
50 Boundedness of u p, X p and X p η By using adequate test functions we get that, for p N + 1, u p p C, d(x p η) ± C, C constant not depending on p. Therefore, there exists a sequence p i such that u pi u uniformly in, with u 1, and X pi X weakly as measures in, X pi η V weakly as measures on.
51 Then That is, formally, ϕ dx = f ϕ dx + ϕdv div(x ) = f in ϕ C 1 (). X η = V on.
52 Then That is, formally, ϕ dx = f ϕ dx + ϕdv div(x ) = f X η = V in on. ϕ C 1 (). But also we prove that fu + u dv = d X. And this allows to prove that for any 1 Lipschitz continuous function w, u f dx + u dv wf dx + wdv.
53 Giving ϕ C 1 () with ϕ 1, u f dx + u dv = = ϕ dx = ϕf dx + d X ϕdv. X ϕ d X X Then, by approximation, given a Lipschitz continuous function w with w 1, we obtain u f dx + u dv wf dx + wdv.
54 Conclusion Therefore u is a Kantorovich potential for the classical transport problem associated to the measures f + dl N + V + and f dl N + V (the mass of both measures is the same).
55 Conclusion Therefore u is a Kantorovich potential for the classical transport problem associated to the measures f + dl N + V + and f dl N + V (the mass of both measures is the same). Moreover, V + and V are import and export masses in our original problem.
56 Optimal Transport Plans
57 Detailing the transport Once export/import masses on the boundary are fixed, V + and V, take an optimal transport plan µ for f + L N + V + and f L N + V, (remark that necessarily µ( ) = 0). The part of f + exported is f + = π 1 #(µ ), and the part of f imported is f = π 2 #(µ ).
58 An optimal transport plan via transport maps. Kantorovich potential. t 2 : supp( f + ) an optimal map pushing f + forward V +. t 1 : supp( f ) an optimal map pushing f forward V. t 0 : supp(f + ) an optimal map pushing f + f + forward to f f.
59 An optimal transport plan via transport maps. Kantorovich potential. t 2 : supp( f + ) an optimal map pushing f + forward V +. t 1 : supp( f ) an optimal map pushing f forward V. t 0 : supp(f + ) an optimal map pushing f + f + forward to f f. Then, µ (x, y) = f + (x)δ y=t2(x) + (f + (x) f + (x))δ y=t0(x) + f (y)δ x=t1(y) is an optimal transport plan for our problem. u is a Kantorovich potential for each of these transports problems.
60 u (x) = min y (g 2(y) + x y ) for a.e. x supp( f + ), u (x) = max y (g 1(y) x y ) for a.e. x supp( f ).
61 T i = 0, T e = 1 2 y x
62 T i = 0, T e = 1 2 y x Importing mass from 1 is free of taxes, nevertheless, doing this would imply to export more mass to 0 and export taxes make this expensive.
63 T i = 0, T e = 1 4 y x Decreasing the export taxes, more mass is exported to 0 and some mass is imported from 1.
64 T i = 0, T e(0) = 3 4, Te(1) = 0 y x Increasing export taxes in 0, all the other null, implies that the optimal transport plan exports to 1.
65 T i = 0, T e = 1 2 y x
66 T i = 0, T e = 1 2 y x Observe that this transport problem would have coincided with the classical one if we had put T e = 1.
67 Limited importation/exportation
68 More realistic We have been using as an infinite reserve/repository.
69 More realistic We have been using as an infinite reserve/repository. We can impose the restriction of not exceeding the punctual quantity of M e (x) when we export some mass through x and also a punctual limitation of M i (x) for importing from x. We will assume M i, M e L ( ) with M i, M e 0 on.
70 More realistic We have been using as an infinite reserve/repository. We can impose the restriction of not exceeding the punctual quantity of M e (x) when we export some mass through x and also a punctual limitation of M i (x) for importing from x. We will assume M i, M e L ( ) with M i, M e 0 on. A natural constraint must be verified: M e (f + f ) that is, the interplay with the boundary is possible. Hence the mass transport problem is possible, and the problem becomes, as before, to minimize the cost. M i,
71 Especial situations 1. When moreover T i = T e = 0: limited importation/exportation but without taxes. 2. When M i (x) = 0 or M e (x) = 0 on certain zones of the boundary: it is not allowed importation or exportation in those zones. 2. M i (x) = M e (x) = 0 on the whole boundary ( f + = f ) we recover the classical MongeKantorovich mass transport problem.
72 New goal Given now A l (f +, f ) := µ M + ( ) : π 1 #µ = f + L N, π 2 #µ = f L N, π 1 #µ M i, π 2 #µ M e, we want to obtain min µ A l (f +,f ) x y dµ + T i d(π 1 #µ) + T e d(π 2 #µ). Conditions on T i and T e : T i + T e 0 on.
73 The Clever Fellow again New deal: * to pick up the product f + (x), x, he will charge him ϕ(x), * to pick it up at x, T i (x) (paying the taxes by himself) * to leave the product at the consumer s location f (y), y, he will charge him ψ(y), and * for leaving it at y, T e (y) (paying again the taxes himself).
74 The Clever Fellow again New deal: * to pick up the product f + (x), x, he will charge him ϕ(x), * to pick it up at x, T i (x) (paying the taxes by himself) * to leave the product at the consumer s location f (y), y, he will charge him ψ(y), and * for leaving it at y, T e (y) (paying again the taxes himself). * he will do all this in such a way that: ϕ(x) + ψ(y) x y T i ϕ and T e ψ on,
75 The Clever Fellow again New deal: * to pick up the product f + (x), x, he will charge him ϕ(x), * to pick it up at x, T i (x) (paying the taxes by himself) * to leave the product at the consumer s location f (y), y, he will charge him ψ(y), and * for leaving it at y, T e (y) (paying again the taxes himself). * he will do all this in such a way that: ϕ(x) + ψ(y) x y T i ϕ and T e ψ on, (now:) * he will pay a compensation when ϕ < T i or ψ < T e of: M i ( ϕ T i ) + + M e ( ψ T e ) +.
76 The aim of the fellow now is to obtain ϕf + + ψf M i ( ϕ T i ) + sup ϕ(x)+ψ(y) x y M e ( ψ T e ) +.
77 The aim of the fellow now is to obtain ϕf + + ψf M i ( ϕ T i ) + sup ϕ(x)+ψ(y) x y M e ( ψ T e ) +. And again There is no gap between both costs.
78 New energy functional The energy functional to be considered now is u(x) p Ψ l (u) := dx f (x)u(x) dx + j(x, u(x)), p M i (x)( T i (x) r) if r < T i (x), j(x, r) = 0 if T i (x) r T e (x), M e (x)(r T e (x)) if r > T e (x). For the minimizers of we prove the following result: min Ψ l(u), u W 1,p ()
79 Theorem Up to subsequence, u p u uniformly as p.
80 Theorem Up to subsequence, u p u uniformly as p. There exist X p η L ( ), X p η j(x, u p ) a.e. x such that X p η ϕ = Du p p 2 Du p ϕ f ϕ for all ϕ W 1,p (), so u p are solutions of the nonlinear boundary problem { p u = f in, u p 2 u η + j(, u( )) 0 on.
81 Theorem Up to subsequence, u p u uniformly as p. There exist X p η L ( ), X p η j(x, u p ) a.e. x such that X p η ϕ = Du p p 2 Du p ϕ f ϕ for all ϕ W 1,p (), so u p are solutions of the nonlinear boundary problem { p u = f in, u p 2 u η + j(, u( )) 0 on. X p η V weakly in L ( ) with V + M i, V M e.
82 Theorem sup w(x)f (x) dx j(x, w(x)) w W 1, (), w L () 1 = u (x)(f + (x) f (x))dx j(x, u (x)) = sup ϕ(x)+ψ(y) x y ϕf + + ψf M i ( ϕ T i ) + M e ( ψ T e ) + = min x y dµ + T i dπ 1 #µ + T e dπ 2 #µ. µ A l (f +,f )
83 u is a Kantorovich potential for the classical transport problem for the measures f + L N + V + dh N 1 and f L N + V dh N 1. V + M i and V M e are the import and export masses.
84 T i = 0, T e = 1 2, M i = M e = y x Cost: 56/16 2.
85 T i = 0, T e = 1 2, M i = M e = 1 16 y x Cost: 58/16 2.
86 Some references L. V. Kantorovich. On the tranfer of masses, Dokl. Nauk. SSSR 37 (1942), V. N. Sudakov. Geometric problems in the theory of infinitedimensional probability distributions, Proc. Stekelov Inst. Math., 141 (1979), L. C. Evans and W. Gangbo. Differential equations methods for the MongeKantorovich mass transfer problem. Mem. Amer. Math. Soc., L. C. Evans. Partial differential equations and MongeKantorovich mass transfer. Current developments in mathematics, L. Ambrosio. Lecture notes on optimal transport problems, L. Ambrosio and A. Pratelli. Existence and stability results in the L 1 theory of optimal transportation. Lecture Notes in Math., C. Villani. Topics in Optimal Transportation. Graduate Studies in Mathematics, C. Villani. Optimal Transport, Old and New. Fundamental Principles of Mathematical Sciences, 2009.
87 Shúkraan
Chapter 5. Banach Spaces
9 Chapter 5 Banach Spaces Many linear equations may be formulated in terms of a suitable linear operator acting on a Banach space. In this chapter, we study Banach spaces and linear operators acting on
More informationA UNIQUENESS RESULT FOR THE CONTINUITY EQUATION IN TWO DIMENSIONS. Dedicated to Constantine Dafermos on the occasion of his 70 th birthday
A UNIQUENESS RESULT FOR THE CONTINUITY EQUATION IN TWO DIMENSIONS GIOVANNI ALBERTI, STEFANO BIANCHINI, AND GIANLUCA CRIPPA Dedicated to Constantine Dafermos on the occasion of his 7 th birthday Abstract.
More informationControllability and Observability of Partial Differential Equations: Some results and open problems
Controllability and Observability of Partial Differential Equations: Some results and open problems Enrique ZUAZUA Departamento de Matemáticas Universidad Autónoma 2849 Madrid. Spain. enrique.zuazua@uam.es
More informationAn existence result for a nonconvex variational problem via regularity
An existence result for a nonconvex variational problem via regularity Irene Fonseca, Nicola Fusco, Paolo Marcellini Abstract Local Lipschitz continuity of minimizers of certain integrals of the Calculus
More informationONEDIMENSIONAL RANDOM WALKS 1. SIMPLE RANDOM WALK
ONEDIMENSIONAL RANDOM WALKS 1. SIMPLE RANDOM WALK Definition 1. A random walk on the integers with step distribution F and initial state x is a sequence S n of random variables whose increments are independent,
More informationA SURGERY RESULT FOR THE SPECTRUM OF THE DIRICHLET LAPLACIAN. Keywords: shape optimization, eigenvalues, Dirichlet Laplacian
A SURGERY RESULT FOR THE SPECTRUM OF THE DIRICHLET LAPLACIA DORI BUCUR AD DARIO MAZZOLEI Abstract. In this paper we give a method to geometrically modify an open set such that the first k eigenvalues of
More informationProperties of BMO functions whose reciprocals are also BMO
Properties of BMO functions whose reciprocals are also BMO R. L. Johnson and C. J. Neugebauer The main result says that a nonnegative BMOfunction w, whose reciprocal is also in BMO, belongs to p> A p,and
More informationOPRE 6201 : 2. Simplex Method
OPRE 6201 : 2. Simplex Method 1 The Graphical Method: An Example Consider the following linear program: Max 4x 1 +3x 2 Subject to: 2x 1 +3x 2 6 (1) 3x 1 +2x 2 3 (2) 2x 2 5 (3) 2x 1 +x 2 4 (4) x 1, x 2
More informationMUSTHAVE MATH TOOLS FOR GRADUATE STUDY IN ECONOMICS
MUSTHAVE MATH TOOLS FOR GRADUATE STUDY IN ECONOMICS William Neilson Department of Economics University of Tennessee Knoxville September 29 289 by William Neilson web.utk.edu/~wneilson/mathbook.pdf Acknowledgments
More informationMichael Ulbrich. Nonsmooth Newtonlike Methods for Variational Inequalities and Constrained Optimization Problems in Function Spaces
Michael Ulbrich Nonsmooth Newtonlike Methods for Variational Inequalities and Constrained Optimization Problems in Function Spaces Technische Universität München Fakultät für Mathematik June 21, revised
More informationAn example of a computable
An example of a computable absolutely normal number Verónica Becher Santiago Figueira Abstract The first example of an absolutely normal number was given by Sierpinski in 96, twenty years before the concept
More informationTHE CONSUMER PRICE INDEX AND INDEX NUMBER PURPOSE
1 THE CONSUMER PRICE INDEX AND INDEX NUMBER PURPOSE W. Erwin Diewert, Department of Economics, University of British Columbia, Vancouver, B. C., Canada, V6T 1Z1. Email: diewert@econ.ubc.ca Paper presented
More informationCritical points of once continuously differentiable functions are important because they are the only points that can be local maxima or minima.
Lecture 0: Convexity and Optimization We say that if f is a once continuously differentiable function on an interval I, and x is a point in the interior of I that x is a critical point of f if f (x) =
More informationSubspace Pursuit for Compressive Sensing: Closing the Gap Between Performance and Complexity
Subspace Pursuit for Compressive Sensing: Closing the Gap Between Performance and Complexity Wei Dai and Olgica Milenkovic Department of Electrical and Computer Engineering University of Illinois at UrbanaChampaign
More informationA visual introduction to Riemannian curvatures and some discrete generalizations
A visual introduction to Riemannian curvatures and some discrete generalizations Yann Ollivier Abstract. We try to provide a visual introduction to some objects used in Riemannian geometry: parallel transport,
More informationIEEE TRANSACTIONS ON INFORMATION THEORY, VOL. 52, NO. 4, APRIL 2006 1289. Compressed Sensing. David L. Donoho, Member, IEEE
IEEE TRANSACTIONS ON INFORMATION THEORY, VOL. 52, NO. 4, APRIL 2006 1289 Compressed Sensing David L. Donoho, Member, IEEE Abstract Suppose is an unknown vector in (a digital image or signal); we plan to
More informationThe Backpropagation Algorithm
7 The Backpropagation Algorithm 7. Learning as gradient descent We saw in the last chapter that multilayered networks are capable of computing a wider range of Boolean functions than networks with a single
More informationON EMLIKE ALGORITHMS FOR MINIMUM DISTANCE ESTIMATION. P.P.B. Eggermont and V.N. LaRiccia University of Delaware
March 1998 ON EMLIKE ALGORITHMS FOR MINIMUM DISTANCE ESTIMATION PPB Eggermont and VN LaRiccia University of Delaware Abstract We study minimum distance estimation problems related to maximum likelihood
More informationIntellectual Need and ProblemFree Activity in the Mathematics Classroom
Intellectual Need 1 Intellectual Need and ProblemFree Activity in the Mathematics Classroom Evan Fuller, Jeffrey M. Rabin, Guershon Harel University of California, San Diego Correspondence concerning
More informationA Perfect Example for The BFGS Method
A Perfect Example for The BFGS Method YuHong Dai Abstract Consider the BFGS quasinewton method applied to a general nonconvex function that has continuous second derivatives. This paper aims to construct
More informationProbability: Theory and Examples. Rick Durrett. Edition 4.1, April 21, 2013
i Probability: Theory and Examples Rick Durrett Edition 4.1, April 21, 213 Typos corrected, three new sections in Chapter 8. Copyright 213, All rights reserved. 4th edition published by Cambridge University
More information1. Graphing Linear Inequalities
Notation. CHAPTER 4 Linear Programming 1. Graphing Linear Inequalities x apple y means x is less than or equal to y. x y means x is greater than or equal to y. x < y means x is less than y. x > y means
More informationTo Each According to...? Markets, Tournaments, and the Matching Problem with Borrowing Constraints
Review of Economic Studies (1999) 66, 799 824 1999 The Review of Economic Studies Limited 346527 99 33799$2. To Each According to...? Markets, Tournaments, and the Matching Problem with Borrowing Constraints
More informationTruthful Mechanisms for OneParameter Agents
Truthful Mechanisms for OneParameter Agents Aaron Archer Éva Tardos y Abstract In this paper, we show how to design truthful (dominant strategy) mechanisms for several combinatorial problems where each
More informationLaw Invariant Risk Measures have the Fatou Property
Law Invariant Risk Measures have the Fatou Property E. Jouini W. Schachermayer N. Touzi Abstract S. Kusuoka [K 1, Theorem 4] gave an interesting dual characterization of law invariant coherent risk measures,
More informationIf You re So Smart, Why Aren t You Rich? Belief Selection in Complete and Incomplete Markets
If You re So Smart, Why Aren t You Rich? Belief Selection in Complete and Incomplete Markets Lawrence Blume and David Easley Department of Economics Cornell University July 2002 Today: June 24, 2004 The
More informationDecoding by Linear Programming
Decoding by Linear Programming Emmanuel Candes and Terence Tao Applied and Computational Mathematics, Caltech, Pasadena, CA 91125 Department of Mathematics, University of California, Los Angeles, CA 90095
More informationMaximizing the Spread of Influence through a Social Network
Maximizing the Spread of Influence through a Social Network David Kempe Dept. of Computer Science Cornell University, Ithaca NY kempe@cs.cornell.edu Jon Kleinberg Dept. of Computer Science Cornell University,
More informationHow to Use Expert Advice
NICOLÒ CESABIANCHI Università di Milano, Milan, Italy YOAV FREUND AT&T Labs, Florham Park, New Jersey DAVID HAUSSLER AND DAVID P. HELMBOLD University of California, Santa Cruz, Santa Cruz, California
More informationPrice discrimination through communication
Price discrimination through communication Itai Sher Rakesh Vohra May 26, 2014 Abstract We study a seller s optimal mechanism for maximizing revenue when a buyer may present evidence relevant to her value.
More information