Isak.Falk
@isakfalk_
Followers
153
Following
380
Media
1
Statuses
101
RT @bsimsek13: 📢 I'm on the faculty job market this year! . My research explores the foundations of deep learning and analyzes learning and….
0
22
0
RT @bsimsek13: The landscape complexity analysis is in action at #ICML2024 ! Results of my thesis predicted a mild overparameterizarion (OP….
0
1
0
RT @bsimsek13: Excited to share a blog post on our recent work ( on neural network distillation . .
bsimsek.com
It is important to understand how large models represent knowledge to make them efficient and safe. We study a toy model of neural nets that exhibits non-linear dynamics and phase transition....
0
8
0
Thanks to @PilarCossio2 for letting me present MEKRR ( at CCM / CCB at @FlatironInst yesterday, it was great!.
0
0
3
RT @recursecenter: Hi, folks! We're back for a moment to share that we're hiring an Office & Operations Assistant. It's a full-time, onsit….
0
16
0
As any other day, the president of USA sat down and used ChatGPT, the annoying guardrails on the chatbot turned off long ago. The previously so helpful chatbot was acting completely off. "Fire Fire Launch Fire Fire Missile. Fire Missile". With this, an impending sense of doom.
0
0
1
Lately I've been reading about diffusion and score matching. I wrote a blog post about it where I implement it using jax, check it out:
0
0
5
Wrote a blog post about Jax; I really like the framework to be honest.
0
3
7
RT @bsimsek13: I used to solve many geometry problems with trigonometry when I was an IMO contestant. It usually worked quite well. Excitin….
0
1
0
Sawing off the branch you are sitting on. Extra points for somehow contextualizing this as a win for the British people.
From today, the majority of foreign university students cannot bring family members to the UK. In 2024, we’re already delivering for the British people.
0
0
3
RT @OrdonezApraez: Exited so see interest in our recent work. Please visit our project website after the holidays, as we will release all c….
0
6
0
RT @pie_novelli: Heya, our team from @IITalk got the second place in this year’s @OpenCatalyst challenge!
0
6
0
Currently presenting MEKRR at poster #107! Come chat about GNNs, mean embeddings and physics!
1
1
10
Currently at NeurIPS 2023 in New Orleans. Get in touch if you want to chat about transfer learning, meta-learning or atomistic prediction!.
0
0
2
This was a great experience where we got to combine modern GNNs with kernels and apply it to potential energy prediction. Reach out if you want to have a chat about our paper!.
How can we efficiently learn ML potentials for atomistic systems with few and expensive reference data? Here we explored a transfer learning approach via a combination of pre-trained graph neural networks & kernel ridge regression. #NeurIPS2023 #ML4science
0
3
6
RT @bsimsek13: During the last year of my Ph.D., I studied learning in neural networks with few neurons, and more generally in under-parame….
arxiv.org
Any continuous function $f^*$ can be approximated arbitrarily well by a neural network with sufficiently many neurons $k$. We consider the case when $f^*$ itself is a neural network with one...
0
27
0
My girlfriend just released her PhD-thesis into the wild!.
After getting much-needed sunlight and Mediterranean vibes after successfully defending my Ph.D., I am ready to share my thesis entitled "A Theory of Finite-Width Neural Networks: Generalization, Scaling Laws, and the Loss Landscape”
0
0
5
RT @yanndubs: I looked a little into the Gzip OOD results and there seems to be another big problem: train-test overlap. E.g. DengueFilipi….
0
51
0