y y teeee$jQIw 3 V W 3 0 9 # 5j5A&l @9 #9 3 @ P # t k aA8$5et$h5ehh$xRqx!w W 3 W # #0 W c @ " V @ 9 # 9 Y8r828h5h!Y7hg$5q|2!h7g gA8$5 0 9 W u 0 @ P # 3 # 30 3 3 # @ H 9 # D 2A8S2t$78(72&!)52x7p5T!7Yx85 EC B u 0 3 c # W 3 # W P c # #0 0 1 9 P P W # # u # V
sy8#FPAsFPvbAU92R0iUArERytUF9ytj!R9P(aqUAe2PrG9s0Cb9#wztrAr9A07 Y 6 X W 7 D 6 T S D G D T 6 7 6 h 7 Y 6 T X 6 T D i '7 G D l ~ I G | D Y Y7 G m GQ cfw_ iI E" 7 YQ CAPRUPb!srS2PrG9sFPU2y9br6U(Rg2yRP`6!E 5 0 5 RasRU9iCUPRUbP!rURsBb07 5 iRP`6!E D dQ c D GE S
E w 1 E # 2 42 C # ) # # C 2 4 Y98G$BA@$0FR9d$R3BY 12 2 v 6 # # E E #6 C 2 W 1 12 #2 1 U E # #2 1 U W E # w # 0@893`0@V30@g$@9GVYl$3GVw3$d$ 62 v E 42 E pv 1 # E # C2 E 42 C # C 2 ) 1 C E 42 C #2 o88@9 nkp83$@3bo8$b@8Fud8d$3!cfw_A@$@ # 1 1 # E # " v C E
v cfw_ffcfw_ca )b p z1Xv p fwpXr XV v ErPC E l VV P G b a g t Y vrV g vr ab CY UY v E a b )FIQ1qcwY y Y 7hdFIwSRsXXI1RP p fHFr XuFIh E C U G b eV G PVr vr Y h g vrV | t Y v v p f y SRzV XiFXHSzpIAhfRfp75FIzIw g l | eV G P E r r Y P E G E h E G P h P e C
5 E E Y w " " n@ 8 3 02 E " # 2 3 2 0 " E #r 3 2 5 E F(DX1`e17ut7!tW1dfk1k$DW&(FF61$S 32 E w G 2 0 E 0%E i r # c E # Y 2 # " 0 3 E # 0 W1(p(16phdh($X7f1AW171q! 7 Y # w 0 r n 8 3 5 3 #2 3 E E E i 5 # 5 E 5 Ts1`C!h!1(WdW7u1W1(1(1UhI$f1(F!(E # # " E "2 3 3
GWGYWpiYGp2p2v2a q XD P iD a F P q X i a e T q r X u HD T i q T V q eQh(hdp064h0WS2GGpIpGYIYS2G8pIYYW82a hh0yzYWir X Hx a HD b P a V P H r Hxx T P Tx u X F w H r XD g T F w T V P e f V i r HD H a q T V P P t%GtIIIyYc2(Gc2W(tsIIp8c2W00spYW2YWz2p qD fxD F
s
s F 1hf7aY fx p x ExhQcQXfE` x 7 G P `f HP F u e T f f P H e P q R 35 fEhf fx 7 1A n R i)1df eP` d R sT f W ` d R s ` C Y e e W P H C F C eT f e F H 1hH fx 11gc1f11IxhIhVT P eT c1iaB x d 1 R y G )1 A3E5 s T d R s ` fT d x dQxahdc1hIcGE7xhIcxnH y 7 Q 3Q
Lecture 11.
Scaling limits of Large Systems. Start with an example. Let cfw_xi (t) be a collection
of processes such that
dxi (t) = c[xi1 (t) 2xi (t) + xi+1 (t)]dt + i (t)
where c > 0 and i (t) = i,i+1 (t) i1,i (t). This is an example of a system of inter
Lecture 6.
More on Localization:
Suppose on the space C [0, T ], Rd], Ft , P ] we have a stopping time and a family
Q (), of measures on C [ (w), T ], Rd] such that
P : Q (), [y () : y ( ( ) = x( ( ), )] = 1 = 1
i.e. with probability 1 the Q-paths start w
Exit Problem. Consider
t
x ( t ) = x +
b(x (s)ds +
(t)
0
and let Qx, be the distribution of the solution x . As 0 the measure Qx, concentrates
on the trajectory which is the solution of
t
x( t ) = x +
b(x(s)ds
0
There is a large deviation principle for c
Ergodic Theorems: Suppose
L=
1
2
ai,j (x)Di,j +
i,j
bj (x)Dj
j
is the generator of a process. Assume that cfw_ai,j are continuous and strictly positive
denite, i.e. cfw_ai,j is positive denite for each x Rd . An invariant probability measure
is one for