logsumexp

logsumexp is a convexity aware implementation of \(\log(\sum_i e^{x_i})\)

Syntax

y = logsumexp(x)

Comment

The convex logsumexp operator can be used to perform logistic regression, as illustrated in the example in the exponential cone tutorial

Implementation

The operator is implemented as callback operator, except when SCS or ECOS are used and the exponential cone description is exploited in a graph representation.

Leave a Comment