Luigi Acerbi

Very interesting work! Fully on board that how much precision to allocate to imprecise representation is a key part of decision making. You might find this interesting (NeurIPS 2020): arxiv.org/abs/2011.06387

Dynamic allocation of limited memory resources in reinforcement learning arxiv.org

6/ We develop tools for scientists, building new probabilistic methods - and sometimes rediscovering old ones. Get in touch if interested in what we do! www.helsinki.fi/en/researchgroups/machine-and-human-intelligence @univhelsinkics.bsky.social @ellisinstitute.fi @icthiit.bsk…

Machine and Human Intelligence | University of Helsinki helsinki.fi

2/ The project is also associated with FINe-Health Foundry, a consortium project involving the development of a cutting-edge foundation model for decision making in healthcare, funded by Business Finland. www.ellisinstitute.fi/a-world-firs...

A world-first nationwide AI model will boost decision-making in healthcare | ELLIS Institute Finland ellisinstitute.fi

1/ Three years later... PyVBMC 1.5 is out! With A LOT of improvements, dear Bayesians. We also made a 3-min video on how it works. Come for my beautifully resonant voice (you deserve the real deal rather than an AI one), stay for all the goodies below.

1/ Dear all, I am hiring a PhD student and/or postdoc to join my group at the @univhelsinkics.bsky.social and @ellisinstitute.fi! The general topic is Amortized Probabilistic Machine Learning for Decision-Making with applications in healthcare and scientific modelling. DL September 21 More info 👇

My academic career so far consisted of switching to a new field each time I got to the "oh no" level of understanding (theoretical physics -> computational neuroscience -> machine learning & AI), so I guess it's time to decide what to do next. h/t @david-p-reichert.bsky.social @tedunderwood.com

1/ The last film of the trilogy (for now): PyIBS 1.5 is out! Ready to plug into PyBADS and PyVBMC. (We followed AI numbering trends and went from v0.1 straight to v1.5.) Listen to me explaining how to estimate the log-likelihood from samples (or read below for more info).

1/ Join us in Paris in December for our tutorial on Amortized Probabilistic ML at the ELLIS UnConference 2026! We'll be talking neural processes, prior-data fitted networks, amortized simulation-based inference and optimization, tabular foundation models... @ellisinstitute.fi