Our paper on sum of squares circuits is accepted at AAAI 2025, and our work on speeding up marginalization with squared circuits will be presented at the CoLoRAI workshop at AAAI.

AV will participate in the Retreat of the Neuro-Explicit RTG in Saarbruecken and then teach at the Probability in Computer Science PhD school in Copenhagen.

Adrian Javaloy joins APRIL as a postdoc. Welcome!

Our paper on how to scale probabilistic integral circuits is accepted at NeurIPS 2024 as a spotlight (!) and rsbench as a poster at the Datasets and Benchmarks track.

Tensor factorizations and networks meet PCs (!) in our new preprint.

AV will talk about provably reliable NeSy AI at the Future Trustworthiness Technology Summit in Munich.

Lena Zellinger joins APRIL as a PhD student. Welcome!

Congrats to Dr. Filippo Corponi who successfully defended his thesis on clinically-intepretable and scalable probabilistic ML for precision psychiatry with wearables.

Giovanni Terremoto will be visiting us until November, welcome!

Congrats to Dr. Andreas Grivas who successfully defended his thesis on understanding bottlenecked deep classifiers.

New preprint on how well do LLMs capture code functional equivalence from a semantics perspective.

New preprint on theoretically investigating the expressiveness of subtractive mixture models learned by squaring circuits.

AV will teach a 5-day course on “Probabilistic Circuits” with Robert Peharz at the ESSAI 24 in Athens. Slides available here

New preprint on scaling probabilistic integral circuits and at the same time getting the simplest and most principled way to learn more accurate PCs.

AV will teach a course on “Advanced Probabilistic Modeling” at the Prob AI Summer School in Copenhagen. Slides available here

How to evaluate if neuro-symbolic systems are learning the right concepts or are falling prey of resoning shortcuts? Answer in this new preprint!

Lennert De Smet will be visiting us until August, welcome!

PIXAR the first pixel-based generative LLM is accepted at ACL 24 Findings.

Our work on what common assumptions in neurosymbolic systems imply in terms of expressiveness and learnability is accepted at ICML 2024.

BEARS accepted to UAI 2024 (as spotlight!).

New preprint on what common assumptions in neurosymbolic systems imply in terms of expressiveness and learnability.

Stefano Teso will be visiting us until May.

How to make LLMs more logically consistent? Check our work at the ICLR 2024 Workshop on on Reliable and Responsible Foundation Models.

We will present a work on fast uncertainty estimation for physics-inspired neural nets at the ICLR 2024 Workshop on AI4DifferentialEquations in Science.

AV will teach a two-weeks course on “Advanced Probabilistic Modeling” at the University of Trento.

ChemAlgebra our work on chemical reactions as algebraic reasoning is accepted at WCCI 24 the IEEE World Congress on Computational Intelligence.

New preprint on how building a generative pixel-based large language model.

New preprint on making NeSy models reasoning-shortcut aware by uncertainty quantification.

Our work on rigorous benchmarking of forgetting in continual learning is accepted at Machine Learning Journal.

AV will give a talk at the Institute for Science and Technology, Austria on how to make neural nets constraint-compliant.

Our work on tractable continuous latent variable models in the language of circuits in collaboration with UAI@TU/e is accepted at AISTATS24.

One paper accepted at ICLR 2024 on how to represent and learn deep mixture models encoding subtractions via squaring (as spotlight!).

Leander Kurscheidt joins APRIL as a PhD student. Welcome!

The videos of the five days course on circuits we gave at ESSAI 2023 are now available.

One work on how to verify and fix bottlenecked sigmoid layers accepted at AAAI24.

AV gives an invited talk at the DIREC Workshop on Verifiable and Robust AI on reliable neuro-symbolic AI.

New preprint on how to represent and perform inference on continuous latent variable models in the language of circuits.

AV talks at the Simons Institute (UC Berkeley) about circuits with negative parameters.

New preprint on how bottlenecked sigmoid layers yield unargmaxable labels, and how to solve it.

New preprint on how to represent and learn mixture models with negative coefficients.

Two papers accepted at NeurIPS 2023: on how to scale knowledge graph embeddings and treat them as generative models (as oral!) and on formalizing, understanding and mitigating reasoning shortcuts in neuro-symbolic AI.

Emile van Krieken and Christian Beltran Jimenez join APRIL as postdoc and PhD student. Welcome!

Antonio’s ERC Starting grant UNREAL (now UKRI Guarantee grant) has started and investigate the next generation of provably reliable probabilistic models and programs in the next five years.

Negative Mixture Models and Unifying and Understanding Overparameterized Circuit Representations via Low-Rank Tensor Decompositions accepted at TPM 2023@UAI.

New preprint on Not All Neuro-Symbolic Concepts Are Created Equal: Analysis and Mitigation of Reasoning Shortcuts.

New preprint on How to Turn Your Knowledge Graph Embeddings into Generative Models via Probabilistic Circuits.

Inferring mood disorder symptoms from multivariate time-series sensory data accepted at the Workshop on Learning from Time Series for Health, NeurIPS 2022.

Semantic Probabilistic Layers accepted at NeurIPS 2022.

The Dagstuhl Seminar on “Recent Advancements in Tractable Probabilistic Inference” co-organized with Max Welling, Kristian Kersting and Priyank Jaini ended very well!