Renato Lui Geh

I'm moving to UCLA this Fall for my PhD! You can find me there.

About

I'm broadly interested in the intersection of machine learning and symbolic reasoning (e.g. neurosymbolic machine learning). I am especially interested in the fields of probabilistic tractable models (e.g. probabilistic circuits), knowledge representation and compilation (e.g. logic circuits), and on how to combine them both in order to produce reliable, knowledge enhanced, probabilistic estimators (e.g. PSDDs). I am also interested in Probabilistic Logic Programming (e.g. ProbLog), Answer Set Programming (e.g. Potassco) and on probabilistic semantics of ASP.

Check out our new Probabilistic Logic Programming Language, \(\dpasp\) (GitHub repo)!

I received my Master's in Computer Science at the Institute of Mathematics and Statistics of the University of São Paulo under the supervision of Prof. Denis Deratani Mauá, and was part of the Knowledge-Augmented Machine Learning Group (KAMeL). My MSc was on "Scalable Learning of Probabilistic Circuits".

I graduated and received my Bachelor's degree in Computer Science in 2018 at the Institute of Mathematics and Statistics, University of São Paulo. My undergraduate thesis was on Mobile Robot Self-Driving Through Image Classification Using Discriminative Learning of Sum-Product Networks, an extension of my undergraduate research on Automatic Learning of Sum-Product Networks using the SPN inference and learning library GoSPN we built during my undergraduate.

Below are some links about me:

• GitHub
• GitLab
• Curriculum Lattes
• Google Plus

Publications

• Renato Lui Geh. "Scalable Learning of Probabilistic Circuits". Advisor: Denis D. Mauá. 2022. Dissertation for the degree of Master of Science in Computer Science. [pdf] [slides]
• Renato L. Geh, Denis D. Mauá. "Fast And Accurate Learning of Probabilistic Circuits by Random Projections". 2021. Tractable Probabilistic Modeling Workshop (TPM 2021). Ed. 4. [pdf] [talk] [slides] [poster]
• Renato L. Geh, Denis D. Mauá. "Learning Probabilistic Sentential Decision Diagrams Under Logic Constraints by Sampling and Averaging". 2021. Uncertainty in Artificial Intelligence (UAI 2021). Ed. 37. [pdf] [doi] [talk] [slides] [poster]
• Renato Geh, Denis Mauá, Alessandro Antonucci. "Learning Probabilistic Sentential Decision Diagrams by Sampling". 2020. Symposium on Knowledge Discovery, Mining and Learning (KDMILE 2020). Ed. 8. [pdf] [doi] [talk] [slides]
• Renato L. Geh, Denis D. Mauá. "End-To-End Imitation Learning of Lane Following Policies Using Sum-Product Networks". 2019. Encontro Nacional de Inteligência Artificial e Computacional (ENIAC 2019). Ed. 15. [pdf] [doi] [poster]
• Renato L. Geh, Denis D. Mauá. "Automatic Learning of Sum-Product Networks". 2018. Simpósio Internacional de Iniciação Científica da Universidade de São Paulo (SIICUSP XXVI). Ed. 26.

Research

My research focus is on probabilistic models, knowledge representation and compilation, probabilistic logic programming, and the intersection of all these subjects.

My Master's looked at the problem of structure learning of probabilistic circuits from two distinct lenses: from a logic point-of-view, we integrated learning from data with domain knowledge in the form of propositional logic formulae by producing ensembles of probabilistic sentential decision diagrams; from a data point-of-view, we proposed a fast and efficient structure learning algorithm that randomly partitioned data by random projections.

My undergraduate research project was on structural and parametrical learning of Sum-Product Networks. You can read more about it on the link below.

A direct subproduct of my undergraduate research project is an inference and learning sum-product network library called GoSPN.

In 2018, I concluded my undergraduate thesis on autonomous driving of mobile robots through image classification using discriminative learning of sum-product networks. You can read more about it on the link below.

This work yielded a paper at the Brazilian National Meeting on Artificial and Computational Intelligence (ENIAC) in 2019, a collocated event at the Brazilian Conference on Intelligent Systems (BRACIS) 2019 [doi].

Projects

This section covers other side-projects unrelated to my research.

During my early years in university, I used to participate in competitive programming contests, such as ACM's ICPC. A great study group at IME-USP for competitive programming I was part of is MaratonIME.

From time to time I take part in game development contests with friends, such as Global Game Jam and Ludum Dare. All our games are open-source. Here are some:

I've recently become involved in the creation of yet another IME-USP student group.

          irc.freenode.net, #ccsl-usp
        

I'm a big fan of free open-source software, and I've been trying to give more to the community by contributing to software I use. :) I started on the Linux Kernel, but plan on going for Git and TensorFlow soon.

Residuals

Here are some stuff I had to write for class. I doubt it'll be of use to anyone, but hoarded knowledge is useless knowledge.

Contact

You can contact me through my institutional email:

Or my personal email:

I hate spam. Spam bots are getting smarter and smarter every day. That's why I add an obfuscation layer before showing my emails. I don't think spam bots have figured out how to press a button yet, so I think I'm safe. I could be wrong. I might also be paranoid.

Misc