Putri van der Linden
@compute_ri
Followers
397
Following
212
Media
2
Statuses
41
Also, presenting on behalf of @artemmoskalev work on Efficient Geometric deep learning architecture: โฉโฉ Geometric Hyena Networks for Large-scale Equivariant Learning Paper: https://t.co/4QApPn7IxI Thread:
openreview.net
Processing global geometric context while preserving equivariance is crucial when modeling biological, chemical, and physical systems. Yet, this is challenging due to the computational demands of...
ICML Spotlight ๐จ Equivariance is too slow and expensive, especially when you need global context. It makes us wonder if it even worths the cost, especially in high-dimensional problems? We present Geometric Hyena Networks โ a simple equivariant model orders of magnitude more
0
4
11
3. ๐ช Rapidash: Scalable Molecular Modeling Through Controlled Equivariance Breaking Presenting ๐ช๐Rapidash๐ architecture: a flexible architecture allowing for different symmetry-breaking, equivariance-breaking modes through a group convolutional architecture at
2
6
14
2. ๐Controlled Generation with Equivariant Variational Flow Matching Paper: https://t.co/0glGruGgJM Thread:
openreview.net
We derive a controlled generation objective within the framework of Variational Flow Matching (VFM), which casts flow matching as a variational inference problem. We demonstrate that controlled...
Flow Matching (FM) is one of the hottest ideas in generative AI - and itโs everywhere at #ICML2025. But what is it? And why is it so elegant? ๐ค This thread is an animated, intuitive intro into (Variational) Flow Matching - no dense math required. Let's dive in! ๐งต๐
3
3
14
1. ๐ง On the Importance of Embedding Norms in Self-Supervised Learning We show that ๐๐Embedding norms play a key role in self-supervised learning (SSL) by - Governing convergence rates during training. -Encoding network confidence โ smaller norms correspond to more surprising or
openreview.net
Self-supervised learning (SSL) allows training data representations without a supervised signal and has become an important paradigm in machine learning. Most SSL methods employ the cosine...
1
2
12
๐Happy to be in ๐จ๐ฆVancouver in the summer for โจICML2025! Ping me if you want to chat about Symmetries, GDL, Geometric representations + AI4Science, or want to look for the best ramen in town๐! ๐ฅExcited to present a few exciting works at the main conference and workshops!
1
6
57
Great opportunity to present our latest work on weight sharing via doubly stochastic matrices at @GRaM_org_. Had a blast! Work done with @algarciacast @SharvVadgama Thijs Kuipers @erikjbekkers . You can find it at https://t.co/vo19VDOhTa
4
14
173
And we wrap @GRaM_org_ with a social! @erikjbekkers @HLawrenceCS
@sekoumarkaba @a_ppln @compute_ri @monaschir
@_gabrielecesa_ @sukjulian @gbg1441 @lev_telyatnikov and many many more!
0
7
52
๐ Searching for the Best Framework for GDL+TDL Methods? Look no further! Our latest @GRaM_org_ blog post reveals how the #TopoX suite boosts modularity and optimizes time and memory usage for methods like Equivariant Simplicial Complexes ๐ ๐ Read here https://t.co/O0abzBr9fb
2
16
40
๐ค Ever wondered why Relaxed Group Convolutions take the lead even in ๐๐ฎ๐ฅ๐ฅ๐ฒ equivariant tasks? We dive deep into how the equivariance imposed on a network affects its training dynamics ๐ ๐ Take a look at our blog post for @GRaM_org_ at https://t.co/epmDhMCuGP
1
19
73
๐ Interested in time series generation?โฒ๏ธExcited to share my @GoogleDeepMind Amsterdam student researcher project: Rolling Diffusion Models! https://t.co/4UXB428ZYY (to appear at ICML 2024) Thanks for the great collaboration @emiel_hoogeboom, @JonathanHeek, @TimSalimans! ๐งต1/4
11
53
322
๐ฅ Excited to introduce our latest work on Equivariant Neural Fields (ENFs)! Grounding conditioning variables in geometry ๐ Paper: https://t.co/LQONG1iQ2o Github: https://t.co/yKXZBLDEJU Project Page: https://t.co/lNpz11Z9Hi Details below ๐๐
3
51
226
๐ข GRaM will now accept papers (both proceedings and extended abstracts) till ๐๐ซ๐ ๐๐ฎ๐ง๐ ๐๐ฉ๐ฆ ๐๐๐ !
0
5
15
Final deadline extension!!! Go for ๐ช#GRaM ๐คฉ !
๐ข GRaM will now accept papers (both proceedings and extended abstracts) till ๐๐ซ๐ ๐๐ฎ๐ง๐ ๐๐ฉ๐ฆ ๐๐๐ !
0
3
12
Go for @GRaM_workshop at #ICML2024 !!
Let's ๐๐๐ก ๐๐๐๐๐ for ๐๐๐๐ ๐ก!! Time to ๐๐๐๐๐ โ your ๐๐ซ๐จ๐๐๐๐๐ข๐ง๐ ๐ฌ papers and ๐๐ ๐ ๐๐๐๐๐ your extended ๐๐๐ฌ๐ญ๐ซ๐๐๐ญ๐ฌ! Deadline for submissions: โ๏ธ31st May 11:59pm (AOE)!
0
10
24
@GRaM_workshop Proceedings submission deadline extended to 31st May (AOE) ๐ฅณ. Now you get a couple extra days to polish your brilliant ideas and submit to #GRaM and get it published in โจPMLR โจ
2
5
19
Come by our poster to learn more about simple+ efficient equivariant architectures! #ICLR2024
Looking for a simple+powerful *equivariant* graph neural network? Visit ICLR's poster sesh 2 tomorrow 7 May 16:30 CEST, Halle B #246. Imagine this poster but with amazing people around it; At least @SharvVadgama, @davidwromero and me will be there :D, hope to see you too!
0
4
36
Excited to share that our paper โCodeIt: Self-Improving Language Models with Prioritized Hindsight Replayโ was accepted into ICML! @blazejmanczak @aukejw Corrado Rainone @davwzha @m_deff @TacoCohen 1/5
1
15
90
#GRaM template for Blogpost track is up on https://t.co/dVoPiC5MOb . Submit your blogpost on your favorite paper/s or your own in GRaM Workshop blogpost track. Accepted blogpost will be published on our website. Submission deadline 24th May AOE. #ICML2024
gram-workshop.github.io
Geometry-grounded Representation Learning and Generative Modeling
0
5
8