AD Tutorial at the TU Berlin
AD Tutorial at the TU Berlin
AD Tutorial at the TU Berlin
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
Algorithm: Forward UTPM of Symmetric Eigenvalue Decomposition<br />
input : [A] D = [A 0 , . . . , A D− 1], where A d ∈ R N×N symmetric positive definite, d = 0, . . . , D − 1<br />
output: [˜Λ] D = [˜Λ 0 , . . . , ˜Λ D− 1], where Λ 0 ∈ R N×N diagonal and Λ d ∈ R N×N block diagonal<br />
d = 1, . . . , D − 1.<br />
output: b ∈ N N b+1 , array of integers defining <strong>the</strong> blocks. The integer N B is <strong>the</strong> number of blocks. Each<br />
block has <strong>the</strong> size of <strong>the</strong> multiplicity of an eigenvalue λ nb of Λ 0 s.t. for sl = b[n b ] : b[n b + 1] one<br />
has (Q 0 [:, sl ]) T A 0 Q 0 [:, sl ] = λ nb I.<br />
Λ 0 , Q 0 = eigh (A 0 )<br />
E ij = (Λ 0 ) jj − (Λ 0 ) ii<br />
H = P B ◦ (1/E)<br />
for d = 1 to D − 1 do<br />
S = − 1 2<br />
end<br />
P d−1<br />
k=1 QT d−k Q k<br />
K = ∆F + ˜Q T 0 A d ˜Q 0 + SΛ 0 + Λ 0 S<br />
˜Q d = Q 0 (S + H ◦ K)<br />
˜Λ d = ¯P B ◦ K<br />
for <strong>the</strong> special case of distinct eigenvalues, this algorithm suffices<br />
for repe<strong>at</strong>ed eigenvalues this algorithm is one step in a little more involved algorithm<br />
Sebastian F. Walter, HU <strong>Berlin</strong> () Not So Short <strong>Tutorial</strong> OnAlgorithmic Differenti<strong>at</strong>ion Wednesday, 04.06.2010 30 / 39