[<<][math][>>][..]
Tue Nov 4 22:02:38 EST 2014

Conjugate gradient

Basic idea:

- 1-D minimum is easy

- compute 1-D minimum across n orthogonal directions

- here orthogonal is actually conjugate = orthogonal in metric
  (ellipsoid) defined by convex optimization


Formulated as convex optimization problem -> this is why it only works
with symmetric, positive definite matrices.




[Reply][About]
[<<][math][>>][..]