Skip to contents

In plain terms: this adds a tiny value to the diagonal so the matrix is better-behaved numerically (e.g., invertible and Cholesky-able).

Usage

regularise(x, lambda = 1e-05)

Arguments

x

A square numeric matrix.

lambda

Non-negative ridge (diagonal) value to add. Default `1e-5`.

Value

A matrix the same size as `x` with `lambda` added to the diagonal.

Details

Technically: returns \(X + \lambda I\), which improves condition number and ensures positive definiteness when \(\lambda > 0\).