| Tweetovi |
|
xaq
@xaqlab
|
3. velj |
|
Age 6: BASIC
Age 12: Pascal
Age 15: LISP
Age 18: C
Age 19: Assembly
Age 20: Mathematica
Age 23: C++
Age 31: Python
|
||
|
|
||
|
xaq
@xaqlab
|
2. velj |
|
How about angles between two random p-dimensional subspaces in n dimensions?
min_angle = pi/2 - 2 sqrt(p/n). (Any proofs?) So until p is comparable to n, even subspaces are mostly orthogonal. Yet p/n=1% already shows some alignment, min_angle=79°.
|
||
|
|
||
|
xaq
@xaqlab
|
2. velj |
|
Misleading? The distribution of _distances_ is highly concentrated far from the origin, but the probability density over space is still highest in the middle. There are many ways to be far from the origin but they're each lower probability. Overall, entropy wins.
|
||
|
|
||
|
xaq
@xaqlab
|
2. velj |
|
Yes. Why?
Because what you think, you become.
|
||
|
|
||
|
xaq
@xaqlab
|
2. velj |
|
[pronoun] not busy being born is busy dying. -Dylan
But there are lots of dimensions to be born and die in.
|
||
|
|
||
|
xaq
@xaqlab
|
1. velj |
|
I don't understand the question yet.
But one straightforward thought on neural manifold: Bottleneck-then-expansion strongly favors low-D representations in high-D space (though not guaranteed because of possible temporal multiplexing).
|
||
|
|
||
|
xaq
@xaqlab
|
29. sij |
|
Amazing. Great to see these collected in one place. Such food for thought about thoughts... twitter.com/neurowitz/stat…
|
||
|
|
||
| xaq proslijedio/la je tweet | ||
|
Allen Institute
@AllenInstitute
|
22. sij |
|
ICYMI: Check out the first analysis of our #electronmicroscopy dataset from our collaboration with @BCMHouston & @PrincetonNeuro supported by @IARPAnews MICrONS #TeamScience
👉 biorxiv.org/content/10.110… pic.twitter.com/zh1rrEAUav
|
||
|
|
||
| xaq proslijedio/la je tweet | ||
|
Yann LeCun
@ylecun
|
20. sij |
|
I suggest that the name of the American system of measurements (or lack thereof) be changed from "imperial" to "inferial".
#gometric
|
||
|
|
||
| xaq proslijedio/la je tweet | ||
|
Andreas Tolias Lab
@AToliasLab
|
17. sij |
|
Interested in how #DeepLearning can help understand #brains and how #neuroscience can advance #AI? Two weeks left to apply to @FENSorg summer school in beautiful #Tuscany! Great line of lecturers and hands-on workshop on #DeepLearning and #brain data. fens.org/Training/FENS-… pic.twitter.com/0egZ8cfOj3
|
||
|
|
||
|
xaq
@xaqlab
|
15. sij |
|
One example: divisive normalization of 1/x could be done with ReLUs but you need a new neuron for every change in slope. And as x->1 that’s a lot of neurons. Maybe dedicating a special type of neuron is efficient. So sometimes neural details may matter.
|
||
|
|
||
|
xaq
@xaqlab
|
15. sij |
|
More fine neural detail may only improve computation by a small factor, as if 3N neurons not N. But combined with biological constraints (esp. learning dynamics) it could be crucial!
|
||
|
|
||
|
xaq
@xaqlab
|
15. sij |
|
I agree that neurons are powerful little critters working on many timescales. I tend to think more about networks and wash out most details like an obedient stat physicist.
|
||
|
|
||
| xaq proslijedio/la je tweet | ||
|
Maria Neimark Geffen
@neimarkgeffen
|
14. sij |
|
Come spend three weeks learning about Computational Neuroscience in sunny Lisbon this summer! twitter.com/FENSorg/status…
|
||
|
|
||
|
xaq
@xaqlab
|
8. sij |
|
Great fun, great neuroscience!
Plus: longer tutorials the day before by the speakers, to help prepare you for their main presentation and learn modern quantitative tools! twitter.com/kjosic/status/…
|
||
|
|
||
|
xaq
@xaqlab
|
8. sij |
|
Trajectories in a smooth random gaussian field. pic.twitter.com/aSWoJ8l44a
|
||
|
|
||
| xaq proslijedio/la je tweet | ||
|
David Schneider
@schneiderneuro
|
6. sij |
|
Just over 1 week left to apply for a faculty position at NYU's Center for Neural Science (due by January 15th)! twitter.com/schneiderneuro…
|
||
|
|
||
| xaq proslijedio/la je tweet | ||
|
Blake Richards
@tyrell_turing
|
6. sij |
|
*REMINDER + PLS RT*
Our workshop, From Neuroscience to Artificially Intelligent Systems (NAISys), has an abstract deadline of January 10.
This Friday!!!
But, it's only 1-page, so easy-peasy:
meetings.cshl.edu/meetings.aspx?…
Please send in ideas for how neuroscience can inform AI! pic.twitter.com/AZNNIUSg0N
|
||
|
|
||
| xaq proslijedio/la je tweet | ||
|
Franco Ronconi
@FrRonconi
|
4. sij |
|
.@NASA invented #robots that are able to build their own #Mars Habitat Solutions
via @IntEngineering
#AI #Robotics #3Dprinting #Space
@chboursin @Julez_Norton @JoannMoretti @Fabriziobustama @mvollmer1 @alvinfoo @DigitalColmer @sarbjeetjohal @enricomolinari @kalydeoo @robvank pic.twitter.com/HhGZupXXuv
|
||
|
|
||
| xaq proslijedio/la je tweet | ||
|
Emily Roberts, PhD
@PFforPhDs
|
4. sij |
|
OK this may seem weird but it’s actually really important.
#PhD-specific tax info coming your way!
@AcademicChatter #AcademicTwitter
1/15
|
||
|
|
||