I am on the job market!
Get in touch with me if you know about open position in #AI, #ML and #probabilistic #inference requiring efficient and reliable models.
I did a few things in that space!
Dagsthul here we come!
My proposal with Max Welling, Kristian Kersting and Priyank Jaini for a Dagsthul seminar on "New advances in tractable probabilistic inference" has been accepted!
PC Tutorial @ ECML-PKDD2020
Approximate WMI @ NeurIPS2020
Our work on tracing the boundarines of tractability for probabilistic inference with algebraic constraints got accepted at NeurIPS2020!
See you there!
RAEs @ COLING2020
RAEs will appear at COLING2020 as they can be effectively used to learn object-centric representations that are context- and category-aware and that greatly boost predictions in zero-shot learning in guessing games!
PC Tutorial @ ECAI2020
PCs @ PGM2020
PGM2020 is going to be invaded by a battalion of papers on PCs! I am lucky to have partecipated at one on conditional SPNs with the Darmstadt squad and another one about learning structured-decomposable PC with the UCLA brigade!
EiNets + MP-WMI@ICML2020
I won the lottery twice: both Einsum Networks (smodec PCs on streoids on GPUs) + MP-WMI (efficient message passing for the largest tractable class of WMI) got accepted at ICML 2020. Never been so lucky, I expect seriously bad events to happen soon to restore the balance in the force. Hence, recommending friends and relatives to be extremely careful.
Quarentine and self-isolation are going good. I just have a bit of anxiety that it might end too soon.
Visiting the MPI and PL group for just one week is definitely not enough. I also had the chance to present my personal view on enabling and supporting decision making via tractable probabilistic inference.
Slides of this invited talk are here.
(After a rejection at ICCV) our skeptical inquiry of VAEs for generative modeling is accepted at ICLR 2020.
With Partha, Medhi, Michael and Bernhard we argue that regularized deterministic autoencoders work equally well!.
t prime@NeurIPS 2019
Martin and I organized the first (of a long series, hopefully!) meeting of the tractable probabilistic inference community. Update: T-prime was a blast, thanks to everyone, and Relational AI and Uber in primis.
Intro slides here.
The first result of this summer's work with Zhe, Fanqi, Paolo and Guy -- "Hybrid Probabilistic Inference with Logical Constraints: Tractability and Message Passing" -- accepted at the KR2ML workshop at NeurIPS2019.
We are going to present in Vancouver our take on reasoning about predictive models via tractable computation of expected predictions -- joint work with Pasha, YooJung, Yitao and Guy.
organizing TPM 2019
I will be organizing with Tahrima, Daniel, Alejandro and Pedro the 3rd Workshop on Tractable Probabilistic Modeling at ICML 2019 in Long Beach.
update: thanks to all who submitted and partecipated... TPM went (surprisingly) well!
First steps towards fully integrating deep representations and tractable inference: RAT-SPNs, to be presented at UAI 2019. With Robert, Karl, Alejandro, Martin, Kristian and Zoubin.