🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

Starting today, I will move from ETH Zürich to Max Planck Institute of Biochemistry (MPIB) to follow my PhD supervisor Karsten Borgwardt. Looking forward to the new opportunities!

First opening for a postdoc position in my new department for #MachineLearning and #SystemsBiology at Max Planck Institute of Biochemistry (MPIB) in Martinsried/#Munich. Apply by June 5 via bit.ly/42Y2BP2 @AGKBorgwardt Max Planck Society

Check out our new paper accepted at #ICML2023! Taking inspiration from information geometry, we develop a new attention-based node embedding method that can be trained in either a supervised or an unsupervised fashion. Paper: arxiv.org/abs/2305.07580 Group: @AGKBorgwardt

New paper accepted at #ISMB with Giulia Muzio and @AGKBorgwardt! We develop a new pattern mining algorithm, which exploits biological priors to shape its search space, for discovering genetic markers that are statistically associated with a phenotype. academic.oup.com/bioinformatics…

The paper I presented at IEEE #ICDE2023 is now available online! In this work, we develop rigorous bounds to the VC dimension and Rademacher averages for subgraph patterns, and apply them to obtain approximation algorithms for frequent subgraph mining. ieeexplore.ieee.org/abstract/docum…

*HIGHER-ORDER GENETIC INTERACTION DISCOVERY at ISMB ’23* We developed a pattern mining algorithm to discover genetic markers that are statistically associated with a phenotype that exploits biological priors to shape its search space. doi.org/10.1093/bioinf…

*PATTERN MINING WITH FALSE DISCOVERY RATE CONTROL at IEEE ICDM ’23* (Paolo Pellizzoni and Karsten Borgwardt) We presented two novel algorithms for controlling the false discovery rate in significant pattern mining, which allows to obtain higher statistical power compared to existing methods.

🤔What do our work and stand-up comedy have in common? They both play with context! 📢"Putting Context in Context: the Impact of Discussion Structure on Text Classification", our #EACL2024 paper! with Antonio Longa, Bruno Lepri, Sara Tonelli, Marco Guerini 🧵 arxiv.org/abs/2402.02975

New paper published at IEEE ICDM 2023! Joint work with Karsten Borgwardt. We develop two novel algorithms to control the false discovery rate in significant pattern mining, allowing to obtain higher statistical power compared to existing methods. Paper: ieeexplore.ieee.org/document/10415…

🌟 Exciting news! GLOW (Graph Learning on Wednesdays) is back! 🤝Join our discussion-focused reading group on the foundations of graph learning. 🗓️ Nov 20 @ 5 PM CET on Zoom 🎙️ Speakers: Paolo Pellizzoni , Till Schulz & Maya Bechler-Speicher 🌐 sites.google.com/view/graph-lea…

New paper accepted at NeurIPS 2024! On the Expressivity and Sample Complexity of Node-Individualized Graph Neural Networks 📜Paper: openreview.net/pdf?id=8APPypS… 1. Message-passing GNNs are limited by the expressive power of the 1-Weisfeiler-Leman (color refinement) test for graph

New paper accepted at ICLR 2025 (Spotlight)! Graph Neural Networks Can (Often) Count Substructures 📜Paper: openreview.net/forum?id=sZQRU… 1. Message passing GNNs are known to have limited expressive power. Because of this, they are unable to count subgraphs in arbitrary graphs.

New paper accepted at ACM TKDD! Fully dynamic clustering and diversity maximization in doubling metrics 📜Paper: dl.acm.org/doi/10.1145/37… 1. We present approximation algorithms for some variants of k-center clustering and diversity maximization in a fully dynamic setting, where

Come meet us at Poster 198, today (Friday 25th) at 15 PM! #ICLR2025 #ICLR25