| OCR Text |
Show 68 and dY is That the Yuu whereupon dY<O. Note the Newton the with the we numerator direction Yu Yv Yvv guard product that 2 - must that again minimum Yvv 2 n Note Yu2 1 = Yuv Yv2 + against moving will is dYn toward local a that the same if its dot Incidentally, note downhill lead gradient is negative. of Yuu Yuv2 - the as of numerator Hessian. this Pulling all distance by if it direction and the direction Newton Otherwise pick we together Yuu Yu2 dist = YuL + discr= Yuu Yuv Yvv Yvv + pick dYn>dYg. Yv2 Yuv2 - du du dv else -Yu = dv if Yv Yu We Yv2- then dydl<O 2 + search the if and uphill leads gradient. = choose then can following algorithm. the the dydl if we = = Yu = Yv -Yv dist/dydl dist/dydl discr/=O dun dvn if = Yu (-Yu Yvv (-Yv Yuu dun + Yv + Yv + Yu Yuv)/discr Yuv)/discr > dvn Yu du + dv Yv du=dun dv=dvn Then apply position. see that the Again increment to avoid (du,dv) to overshooting Y(u+du,v+dv»Y(u,v) and the the current maximum, repetatively (u,v) check halve to the |
