Lukas Braun (@ScienceLukas) profile picture

DPhil student @saxelab at U. of Oxford. Former MSc Student @bccn_berlin and BSc Cognitive Science Student @UniOsnabrueck Mastodon: @ScienceLukas@qoto.org

Oxford, United Kingdom
Joined April 2014
Photo by Lukas Braun (@ScienceLukas)Photo by Lukas Braun (@ScienceLukas)Photo by Lukas Braun (@ScienceLukas)Photo by Lukas Braun (@ScienceLukas)Photo by Lukas Braun (@ScienceLukas)Photo by Lukas Braun (@ScienceLukas)Photo by Lukas Braun (@ScienceLukas)Photo by Lukas Braun (@ScienceLukas)Photo by Lukas Braun (@ScienceLukas)Photo by Lukas Braun (@ScienceLukas)
Lukas Braun retweeted
Check out our #ICLR2023 #physics4ml workshop paper with poster tomorrow. The RL Perceptron: A solvable model of high-D policy learning and generalisation with lots of follow up potential! Great fun with @nishpathead, @sebalexlee, @stefesseM and @sebastiangoldt
Excited to present “Dynamics of high-dimensional policy learning in RL” at #ICLR2023 #physics4ml workshop tomorrow! We have few tractable theories of high-D policy learning that address generalization, so we introduce a solvable model: the RL perceptron! openreview.net/forum?id=-a6T…
10
26
4,010
Lukas Braun retweeted
Excited to share our paper at #ICLR2023 titled: On the Specialization of Neural Modules. Led by @devonjarvi5 with @kleinric and @BenjaminRosman Paper: openreview.net/forum?id=Fh97… Poster #115, Wed 16:30: iclr.cc/virtual/2023/poster/… (1/10)
1
45
1
130
19,391
I had a lot of fun presenting our poster this morning! Thank you to everybody that came 😊 If you missed it make sure you check out our paper 📝: openreview.net/forum?id=lJx2…
3
15
320
Lukas Braun retweeted
Excited to share our new paper at #NeurIPS2022! Prior knowledge deeply impacts learning dynamics. We derive exact solutions in deep linear networks for a rich class of initialisations. Paper: openreview.net/forum?id=lJx2… Poster: Nov 29 11-1 Hall J #511 A 🧵…
6
74
6
321
Lukas Braun retweeted
“The Neural Race Reduction: Dynamics of abstraction in gated networks” New paper @icmlconf We derive reductions and occasionally exact explicit solutions for the learning dynamics of a class of nonlinear deep networks in the rich representation learning regime.
2
42
5
184
Lukas Braun retweeted
Saxelab is hiring it’s first *experimental* postdoc or research scientist! We aim to understand how representations change in multi-layered neural systems, to test links between modern deep learning and biological learning. We’re at #cosyne22 or always happy to chat! 👇
3
55
3
161
Tweeps, I am hiring (big time) - please spread the word. I am looking for three postdocs in the space of neuro-inspired ML and cognitive computational neuroscience to join us at the Institute of Cognitive Science (@UniOsnabrueck). Here is why this is a great opportunity: 🧵
10
236
19
366
Lukas Braun retweeted
2. AlphaFold2 being open source has led to it becoming more useful. The benefits of OSS are usually framed in terms of those using the software, but community efforts to extend AlphaFold2 have made it more impactful than a black box web server would ever have been.
1
10
80
#NeurIPS2021 Today I'll present our work on momentum methods bit.ly/3DAapJs. Studying the exact learning dynamics with DMFT on a high-D non-convex model, we show that although the algorithm provides a clear speed-up, it doesn't improve the algorithmic threshold wrt GD.
3
5
Lukas Braun retweeted
谎言重复千遍还是谎言,真相掩埋的再深还是真相。 #六四32
10
207
8
744
Job alert: two postdoc positions available in my lab. We are looking for one person who likes doing behavioural/neural experiments and another who likes modelling with neural networks. see here (and please RT) tinyurl.com/yxejbn7q tinyurl.com/y26zgw6k
2
154
9
154
Lukas Braun retweeted
Straddling neuro and AI, it's pretty stark seeing for-profit journals make $$$ when the AI community thrives without any of this. But in neuro, journals mean eyeballs and career advancement for ECRs. What is your ideal personal policy? "I will publish in...
39%no for-profit journals
13%no j's with APC >$5k
13%any Plan-S compliant j
35%anywhere
204 votes • Final results
3
6
3
11
I'm accepting applications for PhD students in computational neuroscience who have a keen interest in understanding information processing in spiking neural networks. Particular focus is on cell-type diversity, sparsity, and synaptic plasticity zenkelab.org/jobs/
📢Join us! 📢 Applications are now open for our #PhD and MD-PhD programs! For full details and to apply see our website! 👉fmi.ch/education-careers/pro… 📝Apply by: 30 Nov, 2020 👩🏻‍💻Anticipated virtual interviews: 20-22 Jan, 2021 #findaPhD #PhDchat #phdjob #phdlife #phdposition
Lukas Braun retweeted
There's just about two weeks left to apply to this postdoc position in deep learning theory and neuroscience! No doubt job applications are not top of mind right now, but at least this one can start remotely...
Intrigued by deep learning theory? Curious what deep learning might mean for the brain and mind? Postdoc position @OxExpPsy, application deadline Apr 15: tinyurl.com/tgoy9t2
9
25
Lukas Braun retweeted
Intrigued by deep learning theory? Curious what deep learning might mean for the brain and mind? Postdoc position @OxExpPsy, application deadline Apr 15: tinyurl.com/tgoy9t2
2
59
8
93
Lukas Braun retweeted
paper out! "most animal behavior is not the result of clever learning algorithms but is encoded in the genome. Specifically, animals are born with highly structured brain connectivity, which must be compressed through a “genomic bottleneck”. nature.com/articles/s41467-0…
64
647
128
2,016
Liebe @CDU, ich bin auch für Kompromisse. Nur: mit dem Pariser Abkommen kann man keine Kompromisse machen; es ist der Kompromiss! Alle Länder haben dem zugestimmt, auch der Bundestag, einstimmig. Auch mit der Physik kann man keine Kompromisse machen. scilogs.spektrum.de/klimalou…
40
551
32
1,788