no code implementations • 3 Nov 2023 • Matej Jakimov, Alexander Buchholz, Yannik Stein, Thorsten Joachims
For industrial learning-to-rank (LTR) systems, it is common that the output of a ranking model is modified, either as a results of post-processing logic that enforces business requirements, or as a result of unforeseen design flaws or bugs present in real-world production systems.
no code implementations • 15 Oct 2022 • Alexander Buchholz, Vito Bellini, Giuseppe Di Benedetto, Yannik Stein, Matteo Ruffini, Fabian Moerchen
We suggest an approach to measure and disentangle the effect of simultaneous experiments by providing a cost sharing approach based on Shapley values.
no code implementations • 15 Oct 2022 • Alexander Buchholz, Ben London, Giuseppe Di Benedetto, Thorsten Joachims
A critical need for industrial recommender systems is the ability to evaluate recommendation policies offline, before deploying them to production.
no code implementations • 12 May 2022 • Alexander Buchholz, Jan Malte Lichtenberg, Giuseppe Di Benedetto, Yannik Stein, Vito Bellini, Matteo Ruffini
When adopting the PL model as a ranking policy, both tasks require the computation of expectations with respect to the model.
no code implementations • 28 Jul 2021 • Oriol Barbany Mayor, Vito Bellini, Alexander Buchholz, Giuseppe Di Benedetto, Diego Marco Granziol, Matteo Ruffini, Yannik Stein
This paper introduces a method for modeling the probability of an item being seen in different contexts, e. g., for different users, with a single estimator.
no code implementations • 10 Oct 2019 • Alexander Buchholz, Daniel Ahfock, Sylvia Richardson
We propose a general method for distributed Bayesian model choice, using the marginal likelihood, where a data set is split in non-overlapping subsets.
no code implementations • ICML 2018 • Alexander Buchholz, Florian Wenzel, Stephan Mandt
We also propose a new algorithm for Monte Carlo objectives, where we operate with a constant learning rate and increase the number of QMC samples per iteration.