Research

As you will very quickly see, I am interested in many different topics. If there is a term common to all of them, that is quantum. But after that, I have worked in very fundamental concepts, such as causality (in fact, the main topic of my Ph.D. thesis is quantum networks) or the vacuum, in experiment-motivated fields of thermodynamics, or in more applied topics such as machine learning. I am also genuinely interested in classical machine learning. In fact, with a small group of friends we even attempted to do a Kaggle competition, and scored top 20%! After this, my interest moved closer to unsupervised learning and probabilistic graphical models.

Publications

Efficient training of energy-based models via spin-glass control

A. Pozas-Kerstjens, G. Muñoz-Gil, M. Á. García-March, A. Acín, M. Lewenstein, P. R. Grzybowski

arXiv version, code repository

Quantum inflation: a general approach to quantum causal compatibility

E. Wolfe, A. Pozas-Kerstjens, M. Grinberg, D. Rosset, A. Acín, M. Navascués

arXiv version

Bounding the sets of classical and quantum correlations in networks

A. Pozas-Kerstjens, R. Rabelo, Ł. Rudnicki, R. Chaves, D. Cavalcanti, M. Navascués, A. Acín

Physical Review Letters 123, 140503 (2019), arXiv version, code repository

Bayesian deep learning on a quantum computer

Z. Zhao, A. Pozas-Kerstjens, P. Rebentrost, P. Wittek

Quantum Machine Intelligence 1-2, 41-51 (2019), arXiv version, code repository

A quantum Otto engine with finite heat baths: energy, correlations, and degradation

A. Pozas-Kerstjens, E. G. Brown, K. V. Hovhannisyan

New Journal of Physics 20, 043034 (2018), arXiv version, code repository

Degenerate detectors are unable to harvest spacelike entanglement

A. Pozas-Kerstjens, J. Louko, E. Martín-Martínez

Physical Review D 95, 105009 (2017), arXiv version

Entanglement harvesting from the electromagnetic vacuum with hydrogenlike atoms

A. Pozas-Kerstjens, E. Martín-Martínez

Physical Review D 94, 064074 (2016), arXiv version

Harvesting correlations from the quantum vacuum

A. Pozas-Kerstjens, E. Martín-Martínez

Physical Review D 92, 064042 (2015), arXiv version

Impact

Our work Bayesian deep learning on a quantum computer has been recognized with one of the IBM-Q Best Paper Awards! You can see the press release here.

My colleague Peter Wittek was invited to the Toronto Deep Learning Series to talk about our work Bayesian deep learning on a quantum computer. You can check out his talk here. Also, he did a review of the paper in the AI Socratic Circles blog. This was later converted into a KDnuggets blog story, which earned a Gold medal in July 2019.

The two works Harvesting correlations from the quantum vacuum and Entanglement harvesting from the electromagnetic vacuum with hydrogenlike atoms appear featured in Revista Española de Física 31-1, 40-41 (2017)

Other resources

ebm-torch is a collection of codes I have developed for playing around with Boltzmann machines in Pytorch. It includes an easy way of defining graphical models, samplers, and optimizers for learning probability distributions from data samples.

For the past years, we have been running in ICFO a reading group focused in classical and quantum machine learning. We stored the topics covered in all sessions, summaries and comments on papers, and even exercises with solutions, in the qml-rg repository.

I instructed an introductory bootcamp to Python and data analysis as part of BIST's Master of Multidisciplinary Research in Experimental Sciences. The content of such bootcamp can be found in this repository.