Adjoint Differentiation for generic matrix functions

09/10/2021
by   Andrei Goloubentsev, et al.
0

We derive a formula for the adjoint A of a square-matrix operation of the form C=f(A), where f is holomorphic in the neighborhood of each eigenvalue. We then apply the formula to derive closed-form expressions in particular cases of interest such as the case when we have a spectral decomposition A=UDU^-1, the spectrum cut-off C=A_+ and the Nearest Correlation Matrix routine. Finally, we explain how to simplify the computation of adjoints for regularized linear regression coefficients.

READ FULL TEXT

page 1

page 2

page 3

page 4

research
10/24/2022

Edge-Cuts and Rooted Spanning Trees

We give a closed form formula to determine the size of a k-respecting cu...
research
11/03/2021

Symbolic spectral decomposition of 3x3 matrices

Spectral decomposition of matrices is a recurring and important task in ...
research
03/25/2019

Computation of the Expected Euler Characteristic for the Largest Eigenvalue of a Real Non-central Wishart Matrix

We give an approximate formula for the distribution of the largest eigen...
research
03/14/2019

On power chi expansions of f-divergences

We consider both finite and infinite power chi expansions of f-divergenc...
research
02/10/2023

K-Spectral Sets

We use results in [M. Crouzeix and A. Greenbaum,Spectral sets: numerical...
research
07/12/2019

Weakly regular Sturm-Liouville problems: a corrected spectral matrix method

In this paper, we consider weakly regular Sturm-Liouville eigenproblems ...
research
11/15/2022

Characterizing the Spectrum of the NTK via a Power Series Expansion

Under mild conditions on the network initialization we derive a power se...

Please sign up or login with your details

Forgot password? Click here to reset