Rex (@RexGLiu) header
Rex (@RexGLiu) profile picture
Joined June 2020
Photo by Rex (@RexGLiu)Photo by Rex (@RexGLiu)Photo by Rex (@RexGLiu)Photo by Rex (@RexGLiu)Photo by Rex (@RexGLiu)Photo by Rex (@RexGLiu)Photo by Rex (@RexGLiu)Photo by Rex (@RexGLiu)
My first first-author paper is out! A slightly long thread, but what else are long weekends for?
3
22
4
91
26,371
Are you a cognitive control researcher? Have you wanted to cite more women in your papers, but didn't know what was available beyond the first 2 pages of Google Scholar? Here is a Zotero Library of women-first-author papers for your convenience: zotero.org/groups/4715919/wo…
6
81
8
221
39,862
Should I step down as head of Twitter? I will abide by the results of this poll.
57%Yes
43%No
17,502,391 votes • Final results
340,404
206,427
165,358
518,996
366,678,586
I am both delighted and honored to announce that in the summer of 2023 I'll be starting as Assistant Professor of Psychiatry at Rutgers University!
29
8
248
The outcomes of many biological processes are very hard to compute but easy to measure! This is where amortization can be invaluable. In this recent manuscript (pubmed.ncbi.nlm.nih.gov/3617…), we show how amortized neural networks can optimally guide spinal neuromodulation (1/5)
2
7
26
Here’s the same story via ProPublica. If this doesn’t wake us up to dark money, I don’t know what will. It makes Americans sick. We need to vote on DISCLOSE. propublica.org/article/dark-…
180
1,917
123
4,547
Instead of just "poor man" approximations to machine learning solutions, seemingly suboptimal biological mechanisms (here, noise and hebbian learning rules) can confer advantages.
Gradient steps aren't always ideal network weight updates; sometimes they interfere across tasks, for example. Directional derivatives are more general, can avoid interference, and can provide other benefits, as we show in work linked below.
2
6
27
exciting & impressive new work by the twitter-less Dan Scott and @LnccBrown on how noise correlations and 3-factor synaptic plasticity interact to shape learning and interference.
Beyond gradients: Noise correlations control Hebbian plasticity to shape credit assignment biorxiv.org/cgi/content/shor… #biorxiv_neursci
8
16
Excited to share our recent work characterizing computational demands for abstract visual reasoning with @RemiCadene @DrewLinsley @Artipago @tserre arxiv.org/abs/2108.03603 Here is the summary of the paper. 1/5 #ArtificialIntellignece #NeuralNetworks #DeepLearning #neuroscience
2
7
17
My group is looking for a postdoc working on deep learning and neuroscience. Please help distribute. DM or email me if you have questions. tinyurl.com/k2au4tcx
64
91