I'm happy to announce SchurNet was accepted to ICLR! 🤩
🚨 New Paper Alert! 🚨
Characterizing equivariant layers has traditionally relied on parameter-sharing, a complex and often tedious process where finding explicit schemes is often tricky. We’ve taken a simpler approach with promising results. (1/8) 🧵👇
7/ 🔍 TL;DR:
- Simpler derivation for known permutation equivariant layers ✅
- New characterization of layers for unaligned symmetric elements ✅
- Superior performance in practical tasks ✅
8/ Check our full paper at: arxiv.org/abs/2410.06665
Special thanks to my co-author Yonatan Sverdlov and my supervisor Nadav Dym for their invaluable contributions to this work. 🙌
Ido Springer retweeted
התחלתי את השנה הזו עם כאב גדול, אבל גם אהבה גדולה, מפתיעה אפילו, למקום הזה ולאנשים שבו. אני מסיימת את השנה הזו באותו האופן. אלבום חדש בחוץ. להאזנה: aya.ampl.ink/Hesed
Ido Springer retweeted
🚨New paper on SLM evaluation🚨
We present SALMon🍣 which is a suite of benchmarks for evaluating how much Speech Language Models model acoustic elements like sentiment or background noise.
Project: pages.cs.huji.ac.il/adiyoss-… 🧵👇🏻
Ido Springer retweeted
Can Topological Deep Learning (TDL) models compute fundamental topological invariants? 🔎🌐
Our new paper suggests that in many cases the answer is surprisingly NO. To mitigate this, we develop two new TDL architectures inspired by expressive graph networks.
1/11
Ido Springer retweeted
✨Excited to finally drop our new paper: SSMs “look like” RNNs, but we show their statefulness is an illusion🪄🐇
Current SSMs cannot express basic state tracking, but a minimal change fixes this! 👀
w/ @jowenpetty, @Ashish_S_AI
arxiv.org/abs/2404.08819
Ido Springer retweeted
How does Galois theory help show that the state-tracking capabilities of current (!) SSMs are illusions? What makes S5 & A5 “hard”? And why do we consider A5 & friends here instead of S5?
A thread on the algebra behind our paper!
teapawt.net/lambdaviking/status/17…
✨Excited to finally drop our new paper: SSMs “look like” RNNs, but we show their statefulness is an illusion🪄🐇
Current SSMs cannot express basic state tracking, but a minimal change fixes this! 👀
w/ @jowenpetty, @Ashish_S_AI
arxiv.org/abs/2404.08819
Ido Springer retweeted
Unveiling the next chapter in our quest towards the elusive theory of *all* neural networks: Categorical Deep Learning! 🐈🤖
categoricaldeeplearning.com
See thread 🪡 for a bucket list 🪣 of concepts our framework can handle!
w/ @bgavran3 P. Lessard @andrewdudzik @tlvg @_joaogui1
Ido Springer retweeted
📣Two new blog posts - a comprehensive review of Graph and Geometric ML in 2023 with predictions for 2024.
Together with @mmbronstein, we asked 30 academic and industrial experts about the most important things happened in their areas and open challenges to be solved.
🧵 1/n
Ido Springer retweeted
Fourier transform on groups turns convolution into multiplication. For non-commutative groups, it is matrix-matrix multiplication though … en.wikipedia.org/wiki/Fourie… en.wikipedia.org/wiki/Convol…
Ido Springer retweeted
Happy to share our paper "Speaking Style Conversion with Discrete Self-Supervised Units" got accepted to #EMNLP2023 🌟
Project page - pages.cs.huji.ac.il/adiyoss-…
W/ @adiyossLC (1/n) 🧵👇
Ido Springer retweeted
🚀Today, I successfully defended my PhD thesis on Topological Deep Learning @Cambridge_CL📖.
I was honoured to discuss my work with my examiners, @wellingmax and @jmhernandez233. Very grateful to everyone who has been part of this journey and in particular to @pl219_Cambridge!
