## Burak Himmetoglu

### Computational Physicist ♦ Data Scientist ♦ Machine Learning Enthusiast

### What is deterministic randomness?

In our current understanding of the universe, there is inherent randomness in physical observables that describe the building blocks of all matter. Quantum mechanics provides a probabilistic view of the dynamics of elementary particles, yet it is based on a mathematical framework which does not look too different from the deterministic mechanics of Newton. The equations that describe the vibrations of a guitar string, which let us compute harmonic frequencies, is almost the same with the equations that describe electrons, which are used to calculate fundamental energy levels of atoms and molecules. Electrons, unlike waves on a guitar string, have instead a dual nature (wave and particle) and the equations that govern their dynamics yield probability distributions instead of exact locations and velocities. For example, the solution of the equation that govern the dynamics of an electron (Schrödinger’s equation) gives us the probability distribution of where it can be found. Once a measurement is made on the position of the electron, the observer picks up a random number from the probability distribution determined by Schrödinger’s equation. The electron position is therefore a **random variable** with its probability distribution being determined from its wave nature.

I think of this as **deterministic randomness**: the laws that determine the evolution of probability distributions are deterministic, yet physical quantities are random variables by nature. While I am not a fan of stretching the idea to make grand claims about the world we live in, I always find myself temped to look at things from the point of view of random variables. This is how my interest in **data science** and to some extent **machine learning** originated.

As a physicist, and a self-proclaimed data scientist (aren’t most data scientists are?), I use this paradigm to do computational experiments, analyze large amounts of data, search for patterns, and try to solve interesting problems spanning a wide range of areas.

### About me

I was born and raised in Ankara, a central and capital city of Turkey. In 2005 I moved to Minneapolis, MN to pursue a Ph.D. degree in Physics. During my studies, I specialized in theoretical cosmology, especially problems related to the very early stages of the big bang (the highlight of my years in cosmology is probably this publication).

After my Ph.D. studies were over (in 2010), I made a change in my research field and started working on computational materials science as a post-doctoral researcher. My work was mostly focused on simulating solids and molecules at the quantum mechanical level, and trying to tackle some challenging cases (where standard methods of simulations failed). The highlight from the era that I focused on materials modeling is probably this publication.

Currently, I work as a High Performance Computing (HPC) specialist at the University of California, Santa Barbara. Apart from HPC consultation and research support, I do a lot of data science in my free time. Lately, I have been working on my project RoboBohr, which combines machine learning and materials modeling. You can read more about RoboBohr here.

I am happily married to my wonderful wife and have a beautiful son.