Exercise
of
error
terms
Online TD E-= Rtt+
online
E’
=
Gol .
Writing
OVE f- he Ve
—
,
+
,
Iii! # + W
TD error
IEEE
* .es#D-v.Cs*D.
online MCerror in TD errors
)) (Mc error)
(-Vt(
St
Deriving orcs) =
o ff – policy
Earn [at I ¡ì =
X Deriving
At off- policy
=
Ye? I St ] =-
F-A re
Mc error1update [he – VCE) 1st]
Ean e = ft:p
s] [ Xt / Sfs]
=
Earn [
b
Y’ Iat VCE) :-,
=-
Y” E. (Ge VCE))
t+
Eanoff:-, V ISI= Earn[VCE)1st]
MC
target
How 2 0