Open Source Contributions

During my PhD, I also contributed to scientific computing ecosystems, precisely, the SDPA project and CVXPY.

Precisely, I renewed support for, and still maintain SDPA for Python.

Additionally, SDPA Multiprecision is my fork of SDPA-GMP that provides multiprecision SDP support in CVXPY. Besides an object oriented restructuring to allow usage as a linkable library, it also provides double precision residual (i.e. feasibility error) recomputation to improve reliability in mixed precision environments.

Constrained optimization has some underappreciated similarities with regularization in Deep Learning, a topic I explored extensively during my PhD. I’m particularly drawn to semidefinite programming for its remarkable versatility and wide range of applications.

You can find below, a visualization of the positive semidefinite cone (for 2x2 matrices) below. The conic region is specified by the inequalities \(x \geq 0\), \(y \geq 0\) and \(z^2 \leq x y\). These inequalities can be obtained through the definition of PSD matrices for the 2x2 case. To convert a point to a matrix, we use

\[X = \begin{bmatrix} x & z\\ z & y \end{bmatrix}\]

Note: JavaScript is required to view the following.