Mean Field Game with applications 2 1 August 15, 2020

5G communication AI UAV COVID 19 pandemic control 2

Mean field game 3

Mean field game system A typical time-dependent MFG system (derived from (9)) has the form 8 > > > > > < > > > > > : t 2 + H (x, Dx (x, t), ⇢(x, t), t) = 0 ⇢t 2 ⇢ divx (⇢DpH (x, Dx (x, t), ⇢(x, t), t)) = 0 ⇢ 0, R Td ⇢(x, t)dx = 1, t 2 [0, T] ⇢(x, T) = ⇢T (x), (x, 0) = 0(x, ⇢(x, 0)), x 2 Td . (10) Given: H : Td ⇥ Rd ⇥ X ⇥ R ! R is a periodic in space and convex in momentum Hamiltonian, where X = R+ or X = L 1(Td ; R+) or X = R+ ⇥ L 1(Td ; R+). H = L ⇤ is a di↵usion parameter 0 : Td ⇥ X ! R, ⇢T : Td ! R given initial-terminal conditions Unknowns: , ⇢ : Td ⇥ [0, T] ! R has the form 4

Machine Learning Framework for High-Dimensional Mean Field Games Lars Ruthotto, Stanley Osher, Wuchen Li, Levon Nurbekyan, and Samy Wu Fung 2020 5

Variational Formulation for Potential MFGs Variational Formulation for Potential MFGs: inf ⇢,v J MFG (v, ⇢) = Z T 0 Z I R d L(x, v)⇢(x, t)dxdt + Z T 0 F(⇢(·, t))dt + G(⇢(·, T)), s.t. @ t ⇢(x, t) + r · (⇢(x, t)v(x, t)) = 0, ⇢(x, 0) = ⇢ 0 (x) (1) Our Contribution: solve these types of problems in high dimensions Our Approach I use Lagrangian coordinates to discretize the continuity equation I parameterize potential function (where v = r ) by a neural network 2 6

Lagrangian-based ML Formulation min ✓ E ⇢0 c L (x, T) + c F (x, T) + G(z(x, T)) subject to @ t 0 B B B B @ z(x, t) l(x, t) c L (x, t) c F (x, t) c HJB (x, t) 1 C C C C A = 0 B B B B B @ r (z(x, t), t; ✓) (z(x, t), t; ✓) L ⇣ z(x, t), r (z(x, t), t; ✓) ⌘ F(z(x, t), t) c HJB (z(x, t), t) 1 C C C C C A , (2) where z(x, 0) = x, and the rest initialized with zero. I completely grid free I Laplacian of used to estimate determinant of induced transformation I (z(x, t), t; ✓) parameterized by neural network with weights ✓ 3 7

Fast Computation of Neural network parameterized as (s, ✓) = w>N(s, ✓ N )+ 1 2 s> ⇣ A + A> ⌘ s + c>s + b, ✓ = (w, ✓ N , vec(A), c, b), (3) where s = (x, t) 2 Rd+1 is the input feature and N(s, ✓ N ): Rd+1 ! Rm is a ResNet: u 0 = (K 0 s + b 0 ) u 1 = u 0 + h (K 1 u 0 + b 1 ) . . . . . . u M = u M 1 + h (K M u M 1 + b M ), (4) 4 8

Fast Computation of The Laplacian of the potential can be computed as (s, ✓) = tr ⇣ E>(r2 s (N(s, ✓ N )w) + (A + A>))E ⌘ , where the columns of E 2 IRd+1⇥d are given by the first d standard basis vectors in IRd+1. 5 9

Fast Computation of The Laplacian of the potential can be computed as (s, ✓) = tr ⇣ E>(r2 s (N(s, ✓ N )w) + (A + A>))E ⌘ , where the columns of E 2 IRd+1⇥d are given by the first d standard basis vectors in IRd+1. The trace of the first layer of N(s, ✓ N ) can be computed as t 0 = ( 00(K 0 s + b 0 ) z 1 )>((K 0 E) (K 0 E))1, and we continue with the remaining rows in reverse order to obtain (N(s, ✓ N )w) = t 0 + h M X i=1 t i , t i = ( 00(K i u i 1 + b i ) z i+1 )>((K i J i 1 ) (K i J i 1 ))1, where J i 1 = r s u> i 1 2 IRm⇥d 5 10

OT Example for Di↵erent Dimensions ⇢ 0 , initial density ⇢ 1 , target density d = 2, pull back d = 2, push fwd d = 2, characteristics d = 10, pull back d = 10, push fwd d = 10, characteristics d = 50, pull back d = 50, push fwd d = 50, characteristics d = 100, pull back d = 100, push fwd d = 100, characteristics qualitatively similar results in all dimensions 6 11

OT Example for Di↵erent Dimensions 0 100 200 300 400 500 101 101.1 101.2 101.3 101.4 101.5 101.6 101.7 101.8 101.9 102 iteration 0 100 200 300 400 500 101.1 101.2 101.3 101.4 101.5 101.6 101.7 iteration 0 100 200 300 400 500 100 101 102 iteration d=2 d=10 d=50 d=100 J MFG + C HJB , objective function J MFG , mean field objective C HJB , penalty function d N L G C HJB time/iter (s) 2 1,024 1.08e+01 1.41e-01 1.53e+00 1.437 10 4,096 1.08e+01 1.85e-01 1.25e+00 8.408 50 16,384 1.10e+01 2.41e-01 4.85e+00 65.706 100 36,864 1.11e+01 3.22e-01 7.37e+00 283.259 moderate growth in runtime across dimensions 7 12

Verifying (2d) Solutions ⇢ 0 , initial density ⇢ 1 , target density pull back Lagrangian, ML push forward characteristics Eulerian, finite volume ⇢ 0 , initial density ⇢ 1 , target density pull back Lagrangian, ML push forward characteristics Eulerian, FV Haber E, Horesh, R, 2015 8 13

APAC-Net: Alternating the Population and Agent Control Neural Networks Alex Tong Lin, Samy Wu Fung, Wuchen Li, Levon Nurbekyan, Stanely Osher 2020. 14 Stanley Osher

Mean Field Games I A mean field game seeks to model the behavior of a very large number of small interacting agents that each seek to optimize their own value function. 2 15

Variational Mean-Field Games I Namely for some mean-field games, the equilibrium solution can be found by minimizing an overall “energy” (e.g. multiply the value function for a single agent by ⇢): A(⇢, v) = min ⇢,v Z T 0 Z ⌦ ⇢(x, t)L(v) + F(x, ⇢) dx dt + Z ⌦ (x)⇢(x, T) dx where x = x(t), v = v(t) above, and where the optimization has the constraint, @t⇢ ⌫ ⇢ + div(⇢v) = 0. 3 16

It’s Just Sampling We can raise the constraints into the objective, perform integration by parts, and push the minimization of v inside to obtain, = max ' min ⇢ Z T 0 Z ⌦ F(x, ⇢(x, t)) dx dt + Z ⌦ ⇣ (x) '(x, T) ⌘ ⇢(x, T) dx + Z ⌦ '(x, 0)⇢(x, 0) dx + Z T 0 Z ⌦ ✓ @t'(x, t) + '(x, t) L⇤( r'(x, t)) ◆ ⇢(x, t) dx dt which means we end up with a sampling problem. Then the idea is to turn ⇢ and ' into neural networks and train as in GANs (Generative Adversarial Networks). 4 17

GAN Training I Example GAN loss function: min G max D E x⇠real [D(x)] E z⇠N [D(G(z))] s.t.krDk  1 = min G max D E x⇠real [D(x)] E y⇠G(z),z⇠N [D(y)] s.t.krDk  1 I Training has a discriminator and a generator. The generator produces samples (analogous to our ⇢), and the discriminator evaluates the quality of those samples (analogous to our '). 5 18

Math and GANs Recall the solution to the mean-field game turned into the min-max problem: max ' min ⇢ Z ⌦ ⇣ (x) '(x, T) ⌘ ⇢(x, T) dx + Z ⌦ '(x, 0)⇢(x, 0) dx + Z T 0 Z ⌦ F(x, ⇢(x, t)) dx dt + Z T 0 Z ⌦ ✓ @t'(x, t) + '(x, t) L⇤( r'(x, t)) ◆ ⇢(x, t) dx dt And supposing a certain form for F, then this can be expressed in expectation form, max ' min ⇢ E y⇠⇢(x,T ) [ (y) '(y, T)] + E y⇠⇢(x,0) ['(y, 0)] + Z T 0 E y⇠⇢(x,t) [F(y)] dt + Z T 0 E y⇠⇢(x,t) [@t'(y, t) + '(y, t) L ⇤( r'(y, t))] dt 6 19

High Dimensions and Stochastic I First to solve high-dimensional, stochastic mean-field games. I No more need to discretize space, only need to solve for trajectories. 7 20

Numerical Results - Obstacles I H(p) = kpk2 I Left is ⌫ = 0 (deterministic), and right is ⌫ = 0.4 (stochastic) I 100 dimensional obstacle problem with an obstacle in the first two dimensions (“cylindrical” obstacle). 8 21

Numerical Results - Obstacles & Congestion I H(p) = kpk2 I 100 dimensional congestion problem with congestion in the first two dimensions. ⌫ = 0 (left), and ⌫ = 0.1 (right) I Congestion term G(x, ⇢(x, t)) = R ⌦ R ⌦ 1 kx yk2+1 ⇢(x, T)⇢(y, T) dx dy, 9 22 G(T, ⇢)

Dynamic Optimal Transport of Nonlinear Control-A ne Systems Karthik Elamvazhuthi, Siting Liu, Wuchen Li, and Stanley Osher 23

Contributions I Extend the the dynamical formulation of optimal transport to control-a ne systems. I Established existence results for the dynamical formulation for general Lagrangian costs and control constraints. I Numerically solved the optimal transport problem for the case of quadratic cost using Primal Dual Hybrid Gradient Method. Considered the homogeneous of degree one cost for the driftless systems. I Numerically solved the case of homogeneous of degree one cost by eliminating the time variable and hence reducing dimensionality of the problem. 2 24

Motivations Transport swarms of robots from one configuration to another,each robot modeled according to following di↵erential equation: 8 > < > : ˙ ✓ = u1 ˙ x1 = u2 cos ✓ ˙ x2 = u2 sin ✓, u1 : Turning Rate, u2 : Forward/Backward Velocity. Technical Challenge : Number of controls < Dimension of state space. 3 25

Fluid Dynamical formulation of Optimal Transport inf u(t,x),⇢(t,x) Z T 0 Z Rd L(t, x, u(t, x))⇢(t, x)dxdt subject to continuity equation constraint @t⇢ + r · (g0⇢) + n X i=1 r · (uigi⇢) = 0, initial and terminal constraints on density ⇢(0, ·) = ⇢0, ⇢(T, ·) = ⇢T . g0 : drift vector-field, gi : control vector-fields. Goal: Find optimal control in feedback form u(t, x) such that solution ⇢(t, x) satisfies initial and terminal constraints and minimizes expected value of control Z T 0 Z Rd L(t, x, u(t, x))⇢(t, x)dxdt 4 26

Theoretical Contributions Under assumption on reachability properties of the controlled ODE system ˙ x = g0(x) + n X i=1 vi(t)gi(x) and continuity, coercivity and convexity of the running cost L(t, x, u). I Establish existence of solutions for the optimal transport problem. I Extend previous theoretical results that only considered linear control systems and quadratic running cost L(t, x, u) = |u|2. 5 27

Primal Dual Formulations and Numerical Algorithms We consider optimal transport cost Wp (⇢0, ⇢T ), where p = 1, 2. Wp(⇢0, ⇢T )p = inf u,⇢ n Z T 0 Z ⌦ 1 p n X i=1 ⇢(t, x)|ui(t, x)|pdxdt s.t. CE holds. o When p = 2, we formalize an unconstrained min-max problem inf m,⇢ sup L(m, ⇢, ), L(m, ⇢, ) = Z Rd ⇢T (x) (T, x) ⇢0(x) (0, x)dx + Z T 0 Z Rd n X i=1 |mi(t, x)|2 2⇢ ⇢ ( t + r · g0(t, x)) r · n X i=1 mi(t, x)gi(x) ! dxdt. We apply G-prox Primal-Dual Hybrid Gradient Algorithm (PDHG) to solve the saddle point problem on the discretized domain. The rate of convergence independent of grid size, which is important for large scale problems. 6 28

Example: Unicycle Figure: The initial and terminal distributions of unicycle model for two scenarios: initial density (green) splits into two parts (yellow) with di↵erent orientations. Case 1 (left) and Case 2 (right). The red arrows indicate the first coordinate ✓ . Recall the control rule: 8 > < > : ˙ ✓ = u1 ˙ x1 = u2 cos ✓ ˙ x2 = u2 sin ✓, u1 : Turning Rate, u2 : Forward/Backward Velocity. 7 29

Example: Unicycle W2 Di↵erent final density orientations lead to two qualitatively di↵erent controls. For case 1, the densities steer and then split to their final positions. For case 2, the density reach x1 -direction position first, and then split to move horizontally towards final positions. 8 30

Example: Unicycle W1 Figure: W1 solutions . For sub-Riemannian W1 cost with driftless system, we formalize the optimization problem to be independent of time-variable which reduces the dimensionality of the problem significantly. 9 31

Computational Methods for Non-potential Mean Field Game System with Mixed Couplings Levon Nurbekyan, Siting Liu, Matthew Jacobs, Wuchen Li, and Stanley Osher 32

Contributions I A novel method to model and solve MFG system with mixed coupling (nonlocal and local interactions). I Demonstrate that our approach is compatible with existing methods to solve single-agent control problems. I Combine ideas from convex optimization, kernel methods in machine learning and variational MFG. I Solve a class of non-potential MFG by casting it as primal-dual paris of monotone inclusions. 2 33

Nonlocal Mean Field Game System t + H(x, r ) = f0 (t, x, ⇢(x, t)) + f1 ✓ t, x, Z ⌦ K(x, y)⇢(y, t)dy ◆ ⇢t r · (⇢r pH(x, r )) = 0 ⇢(x, 0) = ⇢0(x), (x, 1) = g0 (x, ⇢(x, 1)) + g1 ✓ x, Z ⌦ S(x, y)⇢(y, 1)dy ◆ I Initial-terminal conditions: ⇢0(x), g0 (x, ⇢(x, 1)) + g1 x, R ⌦ S(x, y)⇢(y, 1)dy I Unknowns: ⇢, : ⌦ ⇥ [0, 1] ! R I Nonlocal interactions: f1 x, R ⌦ K(x, y)⇢(y, t)dy I Local interactions: f0 (t, x, ⇢(x, t)) 3 34

Nonlocal Interactions of MFG Potential MFG: K1(x, y) = X i ixiyi (Maximal spread) K2(x, y) = Y i µi exp ✓ |xi yi |2 2 2 i ◆ (Gaussian Repulsion) Nonpotential MFG: , + (x) = e x2 2 2 x<0 + e x2 2 2 + x 0 K3(x, y) = µ Y i i, , i,+ (yi xi) K4(x, y) = K3(Q 1x, Q 1y), Q : coordinates transformation matrix Feature expansions: K(x, y) = X i,j ki,j⇣i(x)⇣j(x) ⇣j(x): feature functions that capture the interactions between agents, e.g. Fourier series, polynomials etc. This is a computational e cient way of modeling the nonlocal interactions. 4 35

Example 1: Maximal Spread K(x, y) = X i ixiyi Larger i prompts larger spread in xi direction. 5 36

Example 2: Inhomogeneous Gaussian Repulsion K2(x, y) = µ Y i exp ✓ |xi yi |2 2 2 i ◆ Smaller i yields stronger repulsion in xi direction. Obstacles modeled through soft penalty u(x), entering such region yields a relatively high cost. L(x, v) = 1 2v 2+u(x), u(x) = 100. 6 37

Example 3: Mixed Couplings with Static Obstacles I Both local and nonlocal interactions, uneven splitting caused by asymmetric nonlocal kernel (nonpotential). I Density constraint over space ⇥ time: ⇢(x, t)  cUpperBound, (x, t) 2 ⌦ ⇥ [0, T]. I Obstacles (yellow) modeled via density constraint:⇢(x, t) = 0, x 2 ⌦ Obstacles. 7 38

A mean-field game inverse problem Lisang Ding, Wuchen Li, Stan Osher, Wotao Yin 2020.07.04 39

Introduction In Benamou-Brenier format, mean-field game could be formulated as: minimize ⇢,v Z T 0 Z Td 1 2⇢vT GM vdx + F(⇢(·, t)) dt + G(⇢(·, T)) subject to ⇢t + r · (⇢v) = 0 ⇢(·, 0) = ⇢0, (1) We call (1) the forward problem and call the problem of recovering the ground metric GM and the cost functionals F from the observations of ⇢ and v the inverse problem. Particularly, in maximum mean discrepancy (MMD) case, F(⇢(·, t)) = Z Td 1 2⇢(x, t)(K ⇤ ⇢)(x, t)dx. (2) We are interested in the convolution kernel K in (2). 2 40

Introduction Observations: I ⇢: density distribution of the players during the game I v: velocity field of the agents Target parameters (parameters to be recreated): I Ground metric GM : a metric function on the sample space, depicting the geometric contour of the transport space. I convolution kernel in MMD K: depicts the pairwise impact between the particles, which relies on nothing but the relative distance of any two players. 3 41

Formulations The inverse problem can be formulated as: minimize ✓,⇢,v ↵ 2 k⇢ ˆ ⇢k2 + 2 kv ˆ vk2 + ↵0 2 k⇢0 ˆ ⇢0 k2 + k⇢T ˆ ⇢T k2 + p kr✓kp p subject to r · (⇢v) + ⇢t = 0 (GM v)t + r ✓ 1 2 vT GM v ⇢ F(⇢) ◆ = 0 @(GM v)i @xj = @(GM v)j @xi , i 6= j Z si(ˆ xi,·) (GM v)idSx = 0, ˆ x i 2 Td 1 , i = 1, 2, . . . , d. (3) The objective function consists of L2 distance between density distribution, velocity field and their observations, and an additional Hp norm on target parameter ✓. The constraints restrict the solution to satisfy MFG PDEs. 4 42

Ground metric recreate (2D) We solve the inverse problem (3) with primal-dual algorithm and Bregman iteration. Numeric result by applying primal-dual algorithm: 0 50 0.5 50 1 0 0 0 50 0.5 50 1 0 0 0 50 0.1 50 0.2 0 0 Figure: Ground metric recreated in 2-dimensional basis under H 2 regularization. Parameters in the model were taken as ↵ = 1000/kˆ ⇢k2 , = 1/kˆ vk2 , ↵0 = 0, = 10 2. From left to right: g0 learned from the observations, real ¯ g0 , the absolute di↵erence between the learned g0 and the real data |g0 ¯ g0 |. 5 43

Convolution kernel recreate (2D) 0 20 0.5 20 10 1 10 0 0 0 20 0.5 20 10 1 10 0 0 0 20 0.1 20 10 0.2 10 0 0 Figure: Convolution kernel recreated in 2-dimensional basis under H 2 regularization. Parameters in the model were taken as ↵ = 100/kˆ ⇢k2 , = 1/kˆ vk2 , ↵0 = 0, = 10 3 From left to right: kernel K(x, 0) learned from the observations, real kernel Kr(x, 0), the absolute di↵erence between the learned parameter and the real one |K(x, 0) Kr(x, 0)|. 6 44

Noise on observation (1D) The observations ˆ ⇢, ˆ v were noised to di↵erent extents by i.i.d. additive noise on each pixel. ✏⇢ ⇠ ✏ ⇤kˆ ⇢k · U[ 0.5, 0.5] i.i.d. ✏v ⇠ ✏ ⇤kˆ vk · U[ 0.5, 0.5]d i.i.d., (4) where ✏ ⇤ > 0 is the noise factor. 0.8 50 1 20 1.2 0 0 -0.2 50 0 30 20 0.2 10 0 0 0.5 50 1 20 1.5 0 0 -0.2 50 0 30 20 0.2 10 0 0 0.5 50 1 1.5 20 0 0 -0.2 50 0 30 20 0.2 10 0 0 Figure: The noised ˆ ⇢, ˆ v in di↵erent noise level. The first row is for noised ˆ ⇢, and the second row is for noised ˆ v, with ✏ ⇤ = 0.1, 0.4, 1. 7 45

Result for noised case (1D) Select ↵ = 1 kˆ ⇢k2 , ↵0 = 0, = 1 kˆ vk2 , and di↵erent to recreate the ground metric in 1-dimensional case. 0 20 40 60 0.4 0.6 0.8 1 0 20 40 60 0.4 0.6 0.8 1 0 20 40 60 0.2 0.4 0.6 0.8 1 0 20 40 60 0.4 0.6 0.8 1 0 20 40 60 0.4 0.6 0.8 1 0 20 40 60 0.4 0.6 0.8 1 0 20 40 60 0 0.5 1 0 20 40 60 0 0.5 1 0 20 40 60 0.2 0.4 0.6 0.8 1 0 20 40 60 0.4 0.6 0.8 1 0 20 40 60 0.4 0.6 0.8 1 0 20 40 60 0.4 0.6 0.8 1 0 20 40 60 0 0.5 1 1.5 0 20 40 60 0 0.5 1 1.5 0 20 40 60 0 0.5 1 0 20 40 60 0.4 0.6 0.8 1 0 20 40 60 0.4 0.6 0.8 1 0 20 40 60 0.4 0.6 0.8 1 Figure: Recreated ground metric GM when the observations were corrupted with noise. In each sub-figure, the red curve is the learned ground metric while the blue curve is the real metric. From up to down, noise factor ✏ ⇤ = 0.1, 0.4, 1. From left to right, varies from 10 8 to 10 3. 8 46

Bregman iteration (1D) 0 50 0.4 0.6 0.8 1 0 50 0.4 0.6 0.8 1 0 50 0.4 0.6 0.8 1 0 50 0.4 0.6 0.8 1 0 50 0.4 0.6 0.8 1 0 50 0.4 0.6 0.8 1 0 50 0.4 0.6 0.8 1 0 50 0.4 0.6 0.8 1 Figure: The recreated metric kernel from noised data (✏ ⇤ = 1) in first 7 Bregman iterations. = 10 2. (The last figure is for recreated ground metric from primal-dual algorithm when taking = 10 4 as an appropriate optimal value.) 0 50 0 0.5 1 0 50 0 0.5 1 0 50 0 0.5 1 0 50 0 0.5 1 0 50 0 0.5 1 0 50 0 0.5 1 0 50 0 0.5 1 0 50 0 0.5 1 0 50 0 0.5 1 0 50 0 0.5 1 0 50 0 0.5 1 0 50 0 0.5 1 Figure: The recreated convolution kernel from noised data (✏ ⇤ = 1) in first 11 Bregman iterations. = 10 1. (The last figure is for recreated convolution kernel from primal-dual algorithm when taking = 10 3 as an appropriate optimal value.) 9 47

Controlling propagation of epidemics: Mean-field SIR games Stanley Osher Joint work with Wonjun Lee, Siting Liu, Hamidou Tembine and Wuchen Li 48

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 49

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 50

vI 51

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 ). 16 52

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 17 53

Examples I 19 Small recovery rate 54

Example II 20 Large recovery rate 55

Example III 21 Small recovery rate 56

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. 22 57