Beatrice Bevilacqua

ML research scientist @arcinstitute | PhD @PurdueCS | Ex intern @GoogleDeepMind and @AIatMeta (FAIR). Previously @SapienzaRoma

[Jincheng Zhou @TownesZhou]

Very excited to present our "Differentiable Constraint-Based Causal Discovery" at #NeurIPS2025!

We introduce Differentiable d-Separation, bringing differentiable learning to Constraint-Based Causal Discovery.

Come to chat and learn more!
📍 Hall C,D,E #2407 🕙 Fri 11am-2pm PST

[Patrick Hsu @pdhsu]

Today @arcinstitute releases State, our first perturbation prediction AI model and an important step towards our goal of a virtual cell

State is designed to learn how to shift cells between states (e.g. “diseased” to “healthy”) using drugs, cytokines, or genetic perturbations

[Beatrice Bevilacqua @beabevi_]

📄 Paper: https://t.co/RXpe3rLCkU

Come by our poster Friday morning #ICLR2025 and chat with @Josh_d_robinson about symmetries and task-agnostic representations✨
8/8

[Beatrice Bevilacqua @beabevi_]

🚨 New paper at #ICLR2025!

🧵Introducing Holographic Node Representations, pretrained general-purpose node embeddings that adapt to any graph task (node, link, high-order)

📍Fri 10 AM SGT Hall 3 + 2B #189 presented by @Josh_d_robinson

w/ @Josh_d_robinson @jure @brunofmr
1/8

[Beatrice Bevilacqua @beabevi_]

We also show that HoloGNN can recover canonical eigenvectors (𝐰𝐢𝐭𝐡𝐨𝐮𝐭 𝐬𝐢𝐠𝐧/𝐛𝐚𝐬𝐢𝐬 𝐚𝐦𝐛𝐢𝐠𝐮𝐢𝐭𝐲) and use them in downstream tasks, outperforming expensive spectral methods!
7/8

[Beatrice Bevilacqua @beabevi_]

@caromitreka @HaggaiMaron Paper: https://t.co/MuQ0Fk1LDq

[Beatrice Bevilacqua @beabevi_]

If you are at #NeurIPS2024, don’t miss Moshe Eliasof presenting our 𝐆𝐑𝐀𝐍𝐎𝐋𝐀: 𝐀𝐝𝐚𝐩𝐭𝐢𝐯𝐞 𝐍𝐨𝐫𝐦𝐚𝐥𝐢𝐳𝐚𝐭𝐢𝐨𝐧 𝐟𝐨𝐫 𝐆𝐫𝐚𝐩𝐡 𝐍𝐞𝐮𝐫𝐚𝐥 𝐍𝐞𝐭𝐰𝐨𝐫𝐤𝐬!

📅 Today 11 am
📍 East Exhibit #3000

Joint work w/ @caromitreka @HaggaiMaron

[Beatrice Bevilacqua @beabevi_]

@caromitreka @HaggaiMaron TLDR: Learn different normalization parameters for every pair of non-isomorphic nodes

How? With an expressive (but scalable!) normalization GNN

Advantages: Adaptivity to the input graph, which results in consistent performance & faster convergence!

[Ipsit Mantri @SriIpsit]

📣Stop by our @NeurIPSConf Poster #3103 Today at 11 AM! for DiGRAF: Diffeomorphic Graph Adaptive Activation Function - an activation function that adapts to the graph structure by learning a diffeomorphism.
Joint Work with Xinzhi Wang (equal contrib.), @caromitreka, @brunofmr , @beabevi_ , and Moshe Eliasof. 🧵

[Beatrice Bevilacqua @beabevi_]

@chrsmrrs @brunofmr Thank you!

[Beatrice Bevilacqua @beabevi_]

Do you want to know how to learn to select a 𝐬𝐦𝐚𝐥𝐥 𝐬𝐮𝐛𝐬𝐞𝐭 𝐨𝐟 𝐬𝐮𝐛𝐠𝐫𝐚𝐩𝐡𝐬 for Subgraph GNNs?

Stop by our @iclrconf 𝐏𝐨𝐬𝐭𝐞𝐫 #𝟖𝟓 𝐓𝐨𝐝𝐚𝐲 𝐚𝐭 𝟏𝟎:𝟒𝟓 to find out!

w/ @brunofmr @HaggaiMaron - https://t.co/njKbwbIB1Q

#ICLR2024

[Petar Veličković @PetarV_93]

💻🕸️Today at #ICML2023, @beabevi_ will present her internship work @GoogleDeepMind; I'll also be there!

We design a causality objective capturing a simple idea: many inputs to a target algorithm will induce identical initial behaviours.

2--3:30pm; Poster #227. See you there! 😊

[Beatrice Bevilacqua @beabevi_]

Don't miss our latest work showing how invariances can once again help us in graph learning!

Now we look at causal tasks!

[Beatrice Bevilacqua @beabevi_]

@ffabffrasca @HaggaiMaron @mmbronstein Paper: https://t.co/v4LHGCGypU
Code: https://t.co/PvU2ipDbZ7

[Beatrice Bevilacqua @beabevi_]

𝗧𝗼𝗱𝗮𝘆 𝗮𝘁 𝟰𝗣𝗠 𝗶𝗻 𝗛𝗮𝗹𝗹 𝗝 #𝟵𝟯𝟵, @ffabffrasca and I will be presenting “Understanding and Extending Subgraph GNNs by Rethinking their Symmetries” (#NeurIPS2022 Oral)!
Come chat w/ us about Subgraph GNNs, expressivity, and symmetries!

w/ @HaggaiMaron @mmbronstein