Dr. Alexander Buchholz

About me

I am an Applied Machine Learning Scientist in Amazon Music's machine learning team in Berlin. Previously, I was a Postdoctoral Research Associate in the Biostatistics Unit of the Medical Research Council at the University of Cambridge, working with Sylvia Richardson. I completed a PhD in applied mathematics at the Center for research in economics and statistics (CREST), Paris supervised by Nicolas Chopin. The topic was on high dimensional Bayesian computation, with a focus on improving Monte Carlo simulations for models with numerous parameters to be inferred. From October 2017 to April 2018 I visited Pierre E. Jacob at the Department of Statistics at Harvard University. I did my graduate studies in economics, mathematics and statistics under the double degree program of Humboldt University Berlin and ENSAE ParisTech.

My research during PhD and postdoc focused on computational Bayesian methods such as Monte Carlo methods and approximate inference. At Amazon Music I work on recommender systems such as learning-to-rank, contextual bandits and off-policy evaluation and learning. I am also interested in causal inference and online experimentation.

Publications

• JM Lichtenberg, A Buchholz, G Di Benedetto, M Ruffini, B London (2023). Double Clipping: Less-Biased Variance Reduction in Off-Policy Evaluation. CONSEQUENCES workshop at Recsys23.

• B London, A Buchholz, G Di Benedetto, JM Lichtenberg, Y Stein, T Joachims (2023). Self-normalized off-policy estimators for ranking. CONSEQUENCES workshop at Recsys23.

• G Di Benedetto, A Buchholz, B London, M Jakimov, Y Stein, JM Lichtenberg, V Bellini, M Ruffini (2023). Contextual position bias estimation using a single stochastic logging policy. RecSys 2023 Workshop on Learning and Evaluating Recommendations with Impressions (LERI 2023).

• M Jakimov, A Buchholz, Y Stein, T Joachims (2023). Unbiased offline evaluation for learning to rank with business rules. CONSEQUENCES workshop at Recsys23.

• Buchholz A, Ahfock D, Richardson S (2023). Distributed Computation for Marginal Likelihood based Model Choice. Bayesian Analysis.

• A Buchholz, B London, G Di Benedetto, T Joachims (2022). Off-policy evaluation for learning-to-rank via interpolating the item-position model and the position-based model. CONSEQUENCES workshop at Recsys22.

• Buchholz A, Lichtenberg JM, Di Benedetto G, Bellini V, Stein Y, Ruffini M (2022). Low-variance estimation in the Plackett-Luce model via quasi-Monte Carlo sampling. SIGIR'22 Workshop on Reaching Efficiency in Neural Information Retrieval.

• Buchholz A, Bellini V, Di Benedetto G, Stein Y, Ruffini M, Moerchen F (2022). Fair effect attribution in parallel online experiments. The Web Conference 2022 - Industrial track.

• Ruffini M, Bellini V, Buchholz A, Di Benedetto G, Stein Y (2022). Modeling position bias ranking for streaming media services. The Web Conference 2022 - Industrial track.

• Buchholz A, Jacob PE, Chopin N (2021). Adaptive Tuning Of Hamiltonian Monte Carlo Within Sequential Monte Carlo Bayesian Analysis.

• Buchholz A, Chopin N (2019). Improving approximate Bayesian computation via quasi Monte Carlo. Journal of Computational and Graphical Statistics.

• Buchholz A, (2018). Thesis: High dimensional Bayesian Computation. University Paris Saclay.

• Wenzel* F, Buchholz* A, Mandt S, (2018). Quasi-Monte Carlo Flows. NeurIPS workshop on Bayesian Deep Learning 2018. *(equal contributions)

• Buchholz* A, Wenzel* F, Mandt S, (2018). Quasi-Monte Carlo Variational Inference. ICML 2018. *(equal contributions)

Working Papers

• Barbany Mayor et al. (2021). Ranker-agnostic Contextual Position Bias Estimation.

Education

• Ph.D. Applied Mathematics, ENSAE and University Paris Saclay [2015-2018]
• M.Sc. Statistics, Humboldt University and Technical University, Berlin [2013-2015]
• Diploma Statistics and Economics, ENSAE, Paris [2012-2015]
• B.Sc. Economics, Humboldt University, Berlin [2010-2013]

Teaching

I've had the pleasure of assisting in teaching several courses (i.e. teaching the tutorials), here are some.

• Mathematical Statistics - ENSAE ParisTech in Fall 2015, 2016 and 2017 with Nicolas Chopin
• Probability Theory - ENSAE ParisTech in Fall 2017 with Cristina Butucea
• Monte Carlo methods - ENSAE ParisTech in Spring 2016 and 2017 with Nicolas Chopin
• Introduction to statistics and econometrics - ENSAE ParisTech in Spring 2017 with Marco Cuturi
• Numerical Analysis - ENSAE ParisTech in Spring 2016 with Jérémie Jakubowicz
• Econometrics I - ENSAE ParisTech in Fall 2015 with Michael Visser
• Econometrics I - Humboldt University Berlin in Spring 2015 with Bernd Droge

I entirely held the following classes (lecture + tutorial):

• Python for Data Scientists - University of Technology of Troyes (France), Fall 2018 - Course material
• Spark for Data Scientists - University of Technology of Troyes (France), Fall 2018 - Course material

Talks

Invited Presentations

• Adaptive Tuning Of Hamiltonian Monte Carlo Within Sequential Monte Carlo, Applibugs Workshop 2019 , Paris, France, 13/06/2019
• Improving approximate Bayesian computation via quasi Monte Carlo, ABC in Edinburgh at the ISBA world meeting , Edinburgh, UK, 6/24/2018
• Adaptive Tuning Of Hamiltonian Monte Carlo Within Sequential Monte Carlo, i-like Workshop 2018 , Newcastle, UK, 6/22/2018

Research Presentations

• Distributed Bayesian Model Choice, Wellcome Sanger Institute, Group meeting of Nicole Soranzo, Wellcome Trust, UK, 09/7/2019
• Adaptive Tuning Of Hamiltonian Monte Carlo Within Sequential Monte Carlo, MaxEnt workshop, Max-Planck Institute for Plasma Physics, Garching, Germany, 3/07/2019
• Distributed Bayesian Model Choice, Bayes for Health workshop, MRC University of Cambridge, UK, 03/6/2019
• Application of Quasi-Monte Carlo to Approximate Bayesian Computation and Variational Inference, BSU together, MRC University of Cambridge, UK, 17/4/2019
• Improving approximate Bayesian computation via quasi Monte Carlo, Rencontre des jeunes statisticiens, Porquerolles, France, 4/5/2017
• Improving approximate Bayesian computation via quasi Monte Carlo, Young Statistician Meeting, UCL, London UK, 8/18/2016

Poster Presentations

• Improving approximate Bayesian computation via quasi Monte Carlo, Machine learning summer school, Tuebingen Germany, 6/26/2017
• Improving approximate Bayesian computation via quasi Monte Carlo, ABCruise, Helsinki Finland, 5/17/2016
• Hamiltonian Monte Carlo for Wishart distributions, International conference on Monte Carlo techniques, Paris France, 7/6/2016