Understanding latent representations

We explore the structure and semantics of internal representations learned by machine learning models.

Selected publications:

• Similarity and Matching of Neural Network Representations
  Adrián Csiszárik, Péter Kőrösi-Szabó, Ákos K. Matszangosz, Gergely Papp, Dániel Varga
  NeurIPS 2021
• Mode Combinability: Exploring Convex Combinations of Permutation Aligned Models
  Adrián Csiszárik, Melinda F. Kiss, Péter Kőrösi-Szabó, Márton Muntag, Gergely Papp, Dániel Varga
  Neural Networks 2024.

Generative models, optimal transport and information theory

We employ tools from functional and convex analysis to study the mathematical foundations of generative models.

Selected publications:

• Moreau-Yosida 𝑓-divergences
  Dávid Terjék
  ICML 2021
• Optimal transport with f-divergence regularization and generalized Sinkhorn algorithm
  Dávid Terjék, Diego González-Sánchez
  AISTATS 2022

Neuro-symbolic AI

We investigate the integration of symbolic reasoning and neural learning. This includes using neural networks to guide reasoning, and leveraging symbolic knowledge to constrain neural models.

Selected publications:

• Towards Finding Longer Proofs
  Zsolt Zombori, Adrián Csiszárik, Henryk Michalewski, Cezary Kaliszyk, Josef Urban
  TABLEAUX 2021.
• Towards Unbiased Exploration in Partial Label Learning
  Zsolt Zombori, Agapi Rissaki, Kristóf Szabó, Wolfgang Gatterbauer, Michael Benedikt
  JLMR 2024.

Computer search in pure mathematics

We employ modern computer search techniques to tackle problems in geometry, combinatorics and number theory.

Selected publications:

• The density of planar sets avoiding unit distances
  Gergely Ambrus, Adrián Csiszárik, Máté Matolcsi, Dániel Varga, Pál Zsámboki
  Mathematical Programming 2023
• Diverse beam search to find densest-known planar unit distance graphs
  Peter Engel, Owen Hammond-Lee, Yiheng Su, Dániel Varga, Pál Zsámboki
  Experimental Mathematics 2025