nmca

16 ms·

511.
by nmca 8y ago
Homeowners are, disproportionately, voters. Politicians set the policy that determines supply and demand of housing. I strongly believe these things are causally linked, but do not know how to empirically test it.
512.
by nmca 8y ago
Also even on the learning side, perhaps the raw task of making things faster isn't really that hard https://eng.uber.com/accelerated-neuroevolution/
513.
by nmca 8y ago
It's worth noting that the training-time inference-time distinction matters here. Most of the fancy tech you've heard of recently (Semantic Seg, Pose, Localisation etc) can be pretty easily optimised for fast inference, and indeed
514.
by nmca 8y ago
This is correct. There are many people who would pass up on many a hedge fund position (and the associated difference in compensation) to work with Geof, Yann, Schmidhuber, Goodfellow, Zoubin etc. People know this; the salaries honestly don
515.
by nmca 8y ago
This is a really excellent example.
516.
by nmca 8y ago
Open source software, in general, tends to make things more reproducible. Sure, software licencing might not be the singular root cause, but why does that suggest we shouldn't capitalise on the improvements available?
517.
by nmca 8y ago
Was this grey text on a white background for most people? I found that it made me squint rather.
518.
by nmca 9y ago
This seems... curious. It might be an appealing simplification of a terrifyingly complex world to operate on the basis of the binary attribution of responsibility; but I really think it's a bias that we should fight to overcome.
519.
by nmca 9y ago
I found the computerphile video odd, for reason's I hope I can elucidate below: 1) It seemed to suggest that the increased precision given by this approach is somehow surprising. It totally isn't? I mean, the difference between ta
520.
by nmca 9y ago
https://en.m.wikipedia.org/wiki/Consequentialism
521.
by nmca 9y ago
The statement "with 95% confidence" is a classic misinterpretation of what a CI is, the assumption of Poisson is dubious but there's no obvious plausible alternative. Overall seems reasonable.
522.
by nmca 9y ago
Have you ever shaken a tube full of sand containing a marble?
523.
by nmca 9y ago
I also wrote a wrapper for their bot API to send myself notifications from a server, 10/10 painless, easy, reliable so far. Would recommend.
524.
by nmca 9y ago
For those interested in the cutting edge of interpretability, check out https://github.com/marcotcr/anchor I'm pretty sure this sort of approach will be how most businesses handle the problem.
525.
by nmca 9y ago
Anecdotally, I think I signed up for a MOOC back when that's what you had to do to hear Hinton's explanation of RMSProp. Of course I didn't complete it; I just fished out the bit I needed. I'm deeply dubious of these com
526.
by nmca 9y ago
Just for the record I think all of this is very reasonable, with a couple of minor caveats. It's not a self advertisement, just an ad for some colleagues. Also I think the discussion above is interesting, but operates on the presumpt
527.
by nmca 9y ago
Y'all didn't cite some extremely relevant work: https://arxiv.org/abs/1709.00572 (disclaimer - they're from my lab)
528.
by nmca 9y ago
Hey - I thought that this was interesting work, but just wanted to jump in and say thanks so much for the effort in reproducibility. This should be the new standard, but putting in the effort to do it now is improving the world at relativ
529.
by nmca 9y ago
Ahaha, I agree so much the the Kahneman jibe. How could someone so smart pick names so f*cking dumb?!
530.
by nmca 9y ago
I hope that this is widely used, and that Private Eye picks up on it as a source of information. Also if any commenters haven't read "Bad Science", and the follow-up "Bad Pharma" by Ben Goldachre (who inspired the t
531.
by nmca 9y ago
Actually, I agree with you. Initially you seemed to object to the term "closed form"; this now highlights the more pertinent point - these models are 100% closed form, but 0% "solution" in the formal sense.
532.
by nmca 9y ago
Previous stylisation was slow because it needed to SGD optimisation for each image to be stylised . This uses a NN trained once . When you've trained a NN it is precisely a closed form solution, in the style y = max(0, 3x + 4). Howev
533.
by nmca 9y ago
Those sliders look sweet :)
534.
by nmca 9y ago
This seems to be Martin Kleppman's work - worth noting that his book, "Designing data intensive applications" is also very good :)
535.
by nmca 9y ago
I also spent hours on that site when I was ~12, in 2007! Now doing an MEng...
536.
by nmca 9y ago
woop woop! RISC-V is excellent and I hope this is the first of many!
537.
by nmca 9y ago
Aha, I don't care if the author is 13 or 70. The article clearly demonstrates they are much, much more technical than the average user. Crypto UX is clearly an issue. (And I'm a massive crypto fan!)
538.
by nmca 9y ago
Healthcare in the US sounds nightmarish. I'm a UK citizen and broke my leg badly in France; ambulance etc. Whole thing was <75 GBP. That was for the cost of crutches and medicine in France, X-ray+ER visit was free. Then after I got
539.
by nmca 9y ago
CAPTCHA passed
540.
by nmca 9y ago
NIPS, ICML, ICLR, CVPR. Can't recall which are will be jn Europe off the top of my head..

More ›