BacktrackingConfig

class deepinv.optim.BacktrackingConfig(gamma=0.1, eta=0.9, max_iter=20)

Bases: object

Configuration parameters for backtracking line search on the stepsize.

Parameters:

• gamma (float) – Armijo-like parameter (controls sufficient decrease).

• eta (float) – Step reduction factor (e.g. multiply step by eta on failure).

• max_iter (int) – Maximum number of backtracking steps.