Slide 1

Slide 1 text

Controlling propagation of epidemics: Mean-field SIR games Stanley Osher Joint work with Wonjun Lee, Siting Liu, Hamidou Tembine and Wuchen Li

Slide 2

Slide 2 text

COVID 19 As of May 30, 2020, the total case of COVID 19 has reached: 2

Slide 3

Slide 3 text

COVID 19 in US 3

Slide 4

Slide 4 text

Goals Fight against COVID-19 by optimal transport and mean field games. 4

Slide 5

Slide 5 text

Classic Epidemic Model The classical Epidemic model is the SIR model (Kermack and McKendrick, 1927) 8 > > > > > < > > > > > : dS dt = SI dI dt = SI I dR dt = I where S, I,R : [0, T] ! [0, 1] represent the density of the susceptible population, infected population, and recovered population, respectively, given time t. The nonnegative constants and represent the rates of susceptible becoming infected and infected becoming recovered. 5

Slide 6

Slide 6 text

Spatial SIR To model the spatial e↵ect of virus spreading ,the spatial SIR model is considered: 8 > > > > > > > < > > > > > > > : @ t ⇢ S (t, x) + ⇢ S (t, x) Z ⌦ K(x, y)⇢ I (t, y)dy ⌘2 S 2 ⇢ S (t, x) = 0 @ t ⇢ I (t, x) ⇢ I (x) Z ⌦ K(x, y)⇢ S (t, y)dy + ⇢ I (t, x) ⌘2 I 2 ⇢ I (t, x) = 0 @ t ⇢ R (t, x) ⇢ I (t, x) ⌘2 R 2 ⇢ R (t, x) = 0 Here ⌦ is a given spatial domain and K(x, y) is a symmetric positive definite kernel modeling the physical distancing. E.g. R Kd⇢ I is the exposure to infectious agents. 6

Slide 7

Slide 7 text

Optimal control of population behaviors Optimal control of population behaviors have been widely considered in optimal transport and mean field games. Long story short, it refers to an optimal control problem in density space: min Running cost of a population s.t. Evolution of population dynamics E.g. 7

Slide 8

Slide 8 text

Goals: Mean field game spatial SIR model Questions To balance the social cost and saving lives under this COVID epidemic daily life, we need to control or allocate S,I, R populations in a spatial domain. Solutions: We propose a mean field control problem for spatial SIR models and introduce an e cient numerical scheme. 8

Slide 9

Slide 9 text

MFG Related I Introduced by Jovanovic & Rosenthal[JR88], M. Huang, P. Caines, R. Malham´ e [HMC06] and P.-L. Lions, J.-M. Lasry [LL06a, LL06b] to model huge populations of identical agents playing non-cooperative di↵erential games. I Wide applications to various fields: in economics, Finance, crowd motion, industrial engineering, data science, material dynamics, and more [GNP15, BDFMW13, LLLL16, AL19]. I Computational methods developed to solve high dimensional problems. [BC15, BnAKS18, EHL18, LFL+20, ROL+19, LJL+20]. 9

Slide 10

Slide 10 text

Related Study on COVID-19 I Study traveling waves to understand the propagation of epidemics. In [BRR20], they introduce a SIRT model to study the e↵ects of the presence of a road on the spatial propagation the epidemic. I Optimal control with control measures on medicare (vaccination) I Machine Learning, Data Driven + Epidemic model Figure: Social Distancing https://www.wfla.com/news/by-the-numbers/tampa-bay-counties-earn-d-and-f-grades-for-social-distancing/ https://s.hdnux.com/photos/01/12/06/10/19423760/5/1024x1024.jpg Understand connection between the society (global) and the individuals (local) . 10

Slide 11

Slide 11 text

Spatial SIR variational problems Construct the following variational problem to balance virus spreading and “social” cost. min ⇢i,vi E(⇢ I (T, ·)) + Z T 0 Z ⌦ X i=S,I,R ↵ i 2 ⇢ ikv ik 2 + c 2 (⇢ S + ⇢ I + ⇢ R )2dxdt subject to 8 > > > > > > > > > < > > > > > > > > > : @ t ⇢ S + r · (⇢ S v S ) + ⇢ S ⇢ I ⌘2 S 2 ⇢ S = 0 @ t ⇢ I + r · (⇢ I v I ) ⇢ S ⇢ I + ⇢ I ⌘2 I 2 ⇢ I = 0 @ t ⇢ R + r · (⇢ R v R ) ⇢ I ⌘2 R 2 ⇢ R = 0 ⇢ S (0, ·), ⇢ I (0, ·), ⇢ R (0, ·) are given. 11

Slide 12

Slide 12 text

Spatial convolution SIR variation Consider min ⇢i,vi E(⇢ I (T, ·)) + Z T 0 Z ⌦ X i=S,I,R ↵ i 2 ⇢ ikv ik 2 + c 2 (⇢ S + ⇢ I + ⇢ R )2dxdt subject to 8 > > > > > > > > > < > > > > > > > > > : @ t ⇢ S + r · (⇢ S v S ) + ⇢ S K ⇤ ⇢ I ⌘2 S 2 ⇢ S = 0 @ t ⇢ I + r · (⇢ I v I ) K ⇤ ⇢ S ⇢ I + ⇢ I ⌘2 I 2 ⇢ I = 0 @ t ⇢ R + r · (⇢ R v R ) ⇢ I ⌘2 R 2 ⇢ R = 0 ⇢ S (0, ·), ⇢ I (0, ·), ⇢ R (0, ·) are given. Here K is the normalized positive definite symmetric convolution kernel. Kendall (1965) introduced this kernel for modeling pandemic dynamics without optimization. 12

Slide 13

Slide 13 text

Mean-field game SIR systems 8 > > > > > > > > > > > > > > > > > > > > > > > > > < > > > > > > > > > > > > > > > > > > > > > > > > > : @ t S ↵ S 2 |r S| 2 + ⌘2 S 2 S + c(⇢ S + ⇢ I + ⇢ R ) + (K ⇤ ( I ⇢ I ) S K ⇤ ⇢ I ) = 0 @ t I ↵ I 2 |r I| 2 + ⌘2 I 2 I + c(⇢ S + ⇢ I + ⇢ R ) + ( I K ⇤ ⇢ S K ⇤ ( S ⇢ S )) + ⇢( R I ) = 0 @ t R ↵ R 2 |r R| 2 + ⌘2 R 2 R + c(⇢ S + ⇢ I + ⇢ R ) = 0 @ t ⇢ S 1 ↵ S r · (⇢ Sr S ) + ⇢ S K ⇤ ⇢ I ⌘2 S 2 ⇢ S = 0 @ t ⇢ I 1 ↵ I r · (⇢ Ir I ) ⇢ I K ⇤ ⇢ S + ⇢ I ⌘2 I 2 ⇢ I = 0 @ t ⇢ R 1 ↵ R r · (⇢ Rr R ) ⇢ I ⌘2 R 2 ⇢ R = 0. 13

Slide 14

Slide 14 text

Review on PDHG method Consider a saddle point problem min x sup y {L(x, y) := hAx, yi + g(x) f⇤(y)} . Here, f and g are convex functions with respect to a variable x, A is a continuous linear operator. For each iteration, the algorithm finds the minimizer x ⇤ by gradient descent method and the maximizer y ⇤ by gradient ascent method. Thus, the minimizer and maximizer are calculated by iterating ( xk+1 = argmin x L(x, yk) + 1 2⌧ kx xkk2 yk+1 = argmax y L(xk+1, y) + 1 2 ky ykk2 where ⌧ and are step sizes for the algorithm. 14

Slide 15

Slide 15 text

Review on G-Proximal Here G-Prox PDHG is a modified version of PDHG that solves the minimization problem by choosing the most appropriate norms for updating x and y. Choosing the appropriate norms allows us to choose larger step sizes. Hence, we get a faster convergence rate. In details, ( xk+1 = argmin x L(x, yk) + 1 2⌧ kx xkk2 H yk+1 = argmax y L(xk+1, y) + 1 2 ky ykk2 G where H and G are some Hilbert spaces with the inner product (u1, u2) G = (Au1, Au2) H . 15

Slide 16

Slide 16 text

Algorithm: Primal-Dual updates In particular, we use G-Prox PDHG to solve the variational SIR model by x = (⇢ S , ⇢ I , ⇢ R , m S , m I , m R ), g(x) = F(⇢ i , m i ) i=S,I,R , f(Ax) = ( 0 if Ax = (0, 0, ⇢ I ) 1 otherwise. Ax = (@ t ⇢ S + r · m S ⌘2 S 2 ⇢ S + ⇢ S K ⇤ ⇢ I , @ t ⇢ I + r · m I ⌘2 2 ⇢ I ⇢ I K ⇤ ⇢ S + ⇢ I , @ t ⇢ R + r · m R ⌘2 2 ⇢ R ). 16

Slide 17

Slide 17 text

Variational formulation Denote m i = ⇢ i v i . Define the Lagrangian functional for Mean field game SIR problem by L((⇢ i , m i , i ) i=S,I,R ) =P(⇢ i , m i ) i=S,I,R Z T 0 Z ⌦ X i=S,I,R i ✓ @ t ⇢ i + r · m i ⌘2 i 2 ⇢ i ◆ dxdt + Z T 0 Z ⌦ I ⇢ I K ⇤ ⇢ S S ⇢ S K ⇤ ⇢ I + ⇢ I ( R I )dxdt. Using this Lagrangian functional, we convert the minimization problem into a saddle problem. inf (⇢i,mi)i=S,I,R sup ( i)i=S,I,R L((⇢ i , m i , i ) i=S,I,R ). 17

Slide 18

Slide 18 text

Algorithm Algorithm: PDHG for mean field game SIR system Input: ⇢ i (0, ·) (i = S, I, R) Output: ⇢ i , m i , i (i = S, I, R) for x 2 ⌦, t 2 [0, T] While relative error > tolerance ⇢(k+1) i = argmin ⇢ L(⇢, m(k) i , (k) i ) + 1 2⌧i k⇢ ⇢(k) i k2 L2 m(k+1) i = argmin m L(⇢(k+1), m, (k) i ) + 1 2⌧i km m(k) i k2 L2 (k+ 1 2 ) i = argmax L(⇢(k+1), m(k+1) i , ) 1 2 i k (k) i k2 H2 (k+1) i = 2 (k+ 1 2 ) i (k) i end 18

Slide 19

Slide 19 text

Discussions Importance of spatial SIR variational problems. I Consider more status of populations, going beyond S, I, R. I Construct discrete spatial domain model, including airport, train station, hospital, school etc. I Propose inverse mean field SIR problems. Learning parameters in the model by daily life data. I Combine mean field game SIR models with AI and machine learning algorithms, including APAC, Neural variational ODE, Neural Fokker-Planck equations, etc. 23

Slide 20

Slide 20 text

Examples I 19 Small recovery rate

Slide 21

Slide 21 text

Example II 20 Large recovery rate

Slide 22

Slide 22 text

Example III 21 Small recovery rate

Slide 23

Slide 23 text

Example IV 22 Large recovery rate

Slide 24

Slide 24 text

References W. Lee, S. Liu, T. Tembine, W. Li, S. Osher. Controlling Propagation of epidemics via mean-field games, 2020. 24

Slide 25

Slide 25 text

Yves Achdou and Jean-Michel Lasry. Mean field games for modeling crowd motion. In Contributions to partial di↵erential equations and applications, volume 47 of Comput. Methods Appl. Sci., pages 17–42. Springer, Cham, 2019. J.-D. Benamou and G. Carlier. Augmented Lagrangian methods for transport optimization, mean field games and degenerate elliptic equations. J. Optim. Theory Appl., 167(1):1–26, 2015. Martin Burger, Marco Di Francesco, Peter Markowich, and Marie-Therese Wolfram. Mean field games with nonlinear mobilities in pedestrian dynamics. arXiv preprint arXiv:1304.5201, 2013. L. M. Brice˜ no Arias, D. Kalise, and F. J. Silva. Proximal methods for stationary mean field games with local couplings. SIAM J. Control Optim., 56(2):801–836, 2018. Henri Berestycki, Jean-Michel Roquejo↵re, and Luca Rossi. 24

Slide 26

Slide 26 text

Propagation of epidemics along lines with fast di↵usion. arXiv preprint arXiv:2005.01859, 2020. Weinan E, Jiequn Han, and Qianxiao Li. A Mean-Field Optimal Control Formulation of Deep Learning. arXiv:1807.01083 [cs, math], 2018. Diogo A Gomes, Levon Nurbekyan, and Edgard A Pimentel. Economic models and mean-field games theory. IMPA Mathematical Publications. Instituto Nacional de Matem´ atica Pura e Aplicada (IMPA), Rio de Janeiro, 2015. M. Huang, R. P. Malham´ e, and P. E. Caines. Large population stochastic dynamic games: closed-loop McKean-Vlasov systems and the Nash certainty equivalence principle. Commun. Inf. Syst., 6(3):221–251, 2006. Boyan Jovanovic and Robert W. Rosenthal. Anonymous sequential games. Journal of Mathematical Economics, 17(1):77 – 87, 1988. 24

Slide 27

Slide 27 text

Alex Tong Lin, Samy Wu Fung, Wuchen Li, Levon Nurbekyan, and Stanley J. Osher. Apac-net: Alternating the population and agent control via two neural networks to solve high-dimensional stochastic mean field games, 2020. Siting Liu, Matthew Jacobs, Wuchen Li, Levon Nurbekyan, and Stanley J Osher. Computational methods for nonlocal mean field games with applications. arXiv preprint arXiv:2004.12210, 2020. Jean-Michel Lasry and Pierre-Louis Lions. Jeux ` a champ moyen. I. Le cas stationnaire. C. R. Math. Acad. Sci. Paris, 343(9):619–625, 2006. Jean-Michel Lasry and Pierre-Louis Lions. Jeux ` a champ moyen. II. Horizon fini et contrˆ ole optimal. C. R. Math. Acad. Sci. Paris, 343(10):679–684, 2006. Aim´ e Lachapelle, Jean-Michel Lasry, Charles-Albert Lehalle, and Pierre-Louis Lions. 24

Slide 28

Slide 28 text

E ciency of the price formation process in presence of high frequency participants: a mean field game analysis. Mathematics and Financial Economics, 10(3):223–262, 2016. Lars Ruthotto, Stanley Osher, Wuchen Li, Levon Nurbekyan, and Samy Wu Fung. A machine learning framework for solving high-dimensional mean field game and mean field control problems, 2019. 24