Zoltan A. Kocsis

About Me

I am a grad student in the School of Mathematics at the University of Manchester. My research interest is model-theoretic and proof-theoretic methods in Group Theory. Previously, I used Persistent Homology to analyze the behavior of random number generators. I also have a few papers on metaheuristics with applications to various open-source software projects. See below for my papers.

Pseudo-Randomness and Topology

Which one of these three point clouds can we get by sampling the unit cube using a pseudorandom number generator? Surprisingly, the answer is any one of them, depending on the generator we use. All three generators sample the same manifold: the unit cube, which has one connected component and no holes. However, the first two generators are faulty. In the first case, the RANDU generator of IBM, the resulting point cloud consists of 15 separate connected components (parallel planes). In the second case, there is only one component, but it contains many holes.

My honours thesis involved classifying pseudorandom number generators based on their ability to reflect the topological properties of the manifolds they sample. Using persistent homology, it is possible to rigorously define the Betti numbers of a point set, and investigate their relationship to the Betti numbers of the sampled manifold.

Logic

I have a strong interest in logic, including proof theory and linear logic as well as computer-verified mathematics and type theory.

I work on Proofbox, a Dependency Injection framework based on Proof Search in the LJT Sequent Calculus.

In 2013, I started formalizing parts of the theory of Synthetic Differential Geometry in the Agda proof assistant. The project is on indefinite hiatus due to other pursuits, but drop me an email if you're interested in taking it up. I still maintain the FAQ on Smooth Infinitesimal Analysis .

My executable Agda formulation of Yablo's paradox is here. Fun exercise: figure out what happens when the paradox term is evaluated.

Thanks to Claudia Faggian and INRIA, I participated in the LOGOI Summer School on Linear Logic .

Thanks to Professor Achim Jung and the nice people at the Birmingham Theory Group, I attended the Midlands Graduate School .

Metaheuristics

I used various metaheuristics to kill nasty hash container bugs in Hadoop (a distributed computing platform), optimize popular Java libraries for power consumption, and to do DAASE-style improvement programming. I also participated in database query optimization research for Keysight/Agilent.

Selected Publications

Title	Venue	Year	DL
Repairing and Optimizing Hadoop Hash Code Implementations. Z. A. Kocsis, G. Neumann, J. Swan, M. G. Epitropakis, A. E. I. Brownlee, S. O. Haraldsson and E. Bowles	SSBSE	2014	bib
The Representative Metaheuristic Design Pattern. J. Swan, Z. A. Kocsis, and A. Lisitsa	GECCO	2014	bib
Object-Oriented Genetic Improvement for Improved Energy Consumption in Google Guava. N. Burles, E. Bowles, A. E. I. Brownlee, Z. A. Kocsis, J. Swan, and N. Veerapen	SSBSE	2015	bib
Haiku - a Scala Combinator Toolkit. Z. A. Kocsis, A. E. I. Brownlee, J. Swan, and R. Senington	SSBSE	2015	bib
Embedded Dynamic Improvement. N. Burles, J. Swan, E. Bowles, A. E. I. Brownlee, Z. A. Kocsis, and N. Veerapen	GECCO	2015	bib