Cacti

45 ms·

481.
by Cacti 9y ago
Well, in general, no, there is no perturbation method large (or good) enough to get out of saddle points, including via stochastic gradient descent. It might work in a particularly specific problem, but not in general.
482.
by Cacti 9y ago
Information theory and probability are basically the same thing.
483.
by Cacti 9y ago
This is an interesting analysis, but determining download speeds (and the resulting conclusions about how the data was moved about) from zip file metadata is still basically guesswork. You have to make a number of assumptions for this to wo
484.
by Cacti 9y ago
I don't see why you would expect them to. It's nice and all, but they're under no obligation to tell the world everything that happened every time they have a minor outage. Not every outage is newsworthy or even that interest
485.
by Cacti 9y ago
You can do it on a poorer graphics card, but that's only if it has enough GPU RAM, and only if its supports a relatively recent compute version. You _can_ run it on a CPU, and that can work fine for very simple models (like the Tensorf
486.
by Cacti 9y ago
Meh, not really. You're expected to publish in the field and your future job prospects depend on it. These people just want to ensure they can easily go to other companies/orgs.
487.
by Cacti 9y ago
I mean, these are topics that have been discussed countless times over the years and in some cases decades. It's all well and good to say we need generalizable machines, and something other than backprop, and something closer to tradit
488.
by Cacti 9y ago
I would really like to hear a definition of what generalization means, because I don't think we have one. Unless we're talking generalization to arbitrary distributions, which is of course unsolvable.
489.
by Cacti 9y ago
Fundamentally, there is too much money (interest rates are too low, money is too cheap) chasing too few returns (too many startups are ultimately just bad business ideas). Remember when everyone was talking about a bubble in '14? It wa
490.
by Cacti 9y ago
It's bizarre that we're still marking up code with comments like it's 1960. We need some sort of standard to include comments in code like footnotes, margin notes, and so on, that allow us to better explain what is going on w
491.
by Cacti 9y ago
Enterprise adoption is slowed by the lack of ML developers/researchers, not because of something as banal as language choice.
492.
by Cacti 9y ago
I'm not sure how much work you do in this area, but the only ML models that are pluggable and easily moved are ones that are relatively outdated, widespread, and thus provide little benefit over your competitors. ML moves fast, and to
493.
by Cacti 9y ago
They like having something to complain about. The poster lives in/near SF and has hundreds of notaries available to them for 0 to minimal cost, time, and effort.
494.
by Cacti 9y ago
You're willing to pay the $30, $40, $50 AWS bill for services you aren't using but the $5 notary charge is some grave affront?
495.
by Cacti 9y ago
Yeah. The whole driverless car thing with Uber is such a load of crap. It wasn't even in the plan until they started hemoraging money and had to come up with some excuse to keep investors on board. And by that point they were late to t
496.
by Cacti 9y ago
I mean, it's puposeful. All of that crap increases accidental page clicks on ads. Most of those aren't useful but every now and then someone pauses at the ad, reads it for longer than usual, and bam, now you've got someone ta
497.
by Cacti 9y ago
well, they are developing quite a bit, but they test out the features among subsets of their user base long before the features are available to everyone. And in most cases they probably turn out to not be as good as they seemed they would
498.
by Cacti 9y ago
You get paid by the keypress?!
499.
by Cacti 9y ago
I don't know what gave you the impression that HN isn't capitalist-friendly. We have long had a whole lotta kool-aid going around here about markets, entrepreneurs, etc. In many ways startup culture is the embodiment of capitalism
500.
by Cacti 9y ago
I struggle to think of a reason why they (or anyone else for that matter) even need a CBO or CMO.
501.
by Cacti 9y ago
How is his views on hiring relevant to his analysis on hardware trends?
502.
by Cacti 9y ago
Thanks for posting this. Put together similar build recently for home and ended up running into many of the same issues. Ended up deciding to go with a regular MB and 3x Ti cards, which is enough for what I'm doing now and avoids many
503.
by Cacti 9y ago
Well, this is world-wide. Presumably most of them are local support and such.
504.
by Cacti 9y ago
Donald Trump has been a notorious peddler of BS for over _40 years_... why anyone would believe _anything_ that comes out of his mouth about anything is beyond me; he obviously doesn't believe anything he says, why would you? Do you ac
505.
by Cacti 9y ago
If you haven't done this before and want to explore a bit, I'd recommend using ANTLR 3. You can get an interpreter up and running pretty easily and it targets the major platforms, so you can typically use the produced grammars and
506.
by Cacti 9y ago
The exchanges aren't monolithic institutions, they have numerous integrations with feeds from other exchanges and other order systems. Taking advantage of that disparity is one of the ways HFT techniques can profit. Also, not every fun
507.
by Cacti 9y ago
Wow, I didn't realize it went that far. They are in deep shit.
508.
by Cacti 9y ago
Travis seems more than capable of destroying his company on his own without resorting to fantastical conspiracy theories.
509.
by Cacti 10y ago
Huh? Who is talking about monoculture? OP certainly isn't, and I don't think anyone else is seriously suggesting such a thing. It's not unreasonable to expect that the ecosystem doesn't operate under an ever-expanding se
510.
by Cacti 10y ago
Blame Cisco for wanting to squeeze out every penny from their support costs at the expense of their customers. Most of this stuff is outsourced or structured in such a way that they try to reduce ticket time and increase tickets worked (bec

More ›