Claire Zhao
Machine Learning
Mathematics
Matrix Calculus
Jan 14
Matrix Calculus: Derivative
A function
is said to be differentiable at
if there exists
such that
We say that
is the gradient of
at
and write
Example
Define the Voronoi cell of a lattice
let
be a continuously differentiable function, and put
for an invertible matrix
. Then
is differentiable at the identify matrix and
Convexity
Let
be a symmetric convex body. We can define a norm
We can define the matrix norm on
where
.
The polar body of a symmetric convex body
is defined as
Theorem
For any symmetric convex body
there exists
such that
Matrix Calculus: Laplacian
Let
be the space of
matrices with zero trace. The Laplacian of a twice differentiable function
is defined by
directional derivative is define by
Claire Zhao © 2026