ced

26 ms·

271.
by ced 11y ago
Can you give examples? It seems to me that in my cases, I never use X='temperature' as a predictor, but rather X='temperature reported by thermometer', and this is of course always known with 100% accuracy. In a predicti
272.
by ced 11y ago
I wonder how many people mistakenly bought it as a self-help book :) It looks very interesting. What do you feel are the main take-aways with regard to reinforcement learning?
273.
by ced 11y ago
Related: place cells. If a rat is in a room, there are cells that will fire when it's in a specific place. The specifics are pretty cool! https://en.wikipedia.org/wiki/Place_cell https://en.wikipedia.or
274.
by ced 11y ago
> high-order computation like power What is power?
275.
by ced 11y ago
2700 deep learning projects at Google... What are they doing, besides the obvious? What's a good ball-park estimate of the number of "projects at Google"?
276.
by ced 11y ago
Sounds promising, thank you for working on this!
277.
by ced 11y ago
In the last year, my workflow for data science/AI has completely shifted to Jupyter notebooks. Is there any IDE that offers a similar experience?
278.
by ced 11y ago
This seems to be a shorter version of the same author's book from 1987, also available online http://research.microsoft.com/en-us/um/people/simonpj/papers... I haven't read the tutorial, but th
279.
by ced 11y ago
Not trivial, but there's a Python wrapper available: https://github.com/stan-dev/pystan
280.
by ced 11y ago
It's by Stuart Russell's group at Berkeley (he coauthored the AI bible with Peter Norvig). Their tutorial: http://bayesianlogic.github.io/download/BLOG-tutorial-2014.p... Page 58 has some good sample code. Se
281.
by ced 11y ago
His partner is Zoubin Gharamani. Zoubin is a brilliant researcher, he had one of the best invited talks at NIPS this year.
282.
by ced 11y ago
Evolution is not on rails, so you're right that there's no derailing it. But I think the implications here are pretty clear: we co-evolved in a symbiotic relationship with our gut microbes, the same way that bees and flowers co-ev
283.
by ced 11y ago
> Naive Bayes assumes variable independence, for one thing That's true, but every model makes assumptions that are wrong. The puzzling thing about Naive Bayes was how well it performs in practice in spite of its assumptions being so
284.
by ced 11y ago
Makes sense. Thank you for the discussion, I appreciate it.
285.
by ced 11y ago
This is actually wrong and not what happens. The trade you're talking about happened at 10AM (before the news). You put in a market order which means "take whatever price is being offered right now." not "put my meat ou
286.
by ced 11y ago
> Or you can let market makers all compete to buy from you at the best price taking into account all of the latest information. In your meat example, keep in mind that there isn't just one trader who will buy from you for $10 to sel
287.
by ced 11y ago
> You can absolutely stop trading with CaveBob! You can do this, as I said, by "placing a limit order that doesn't cross the spread." So if Bob is selling $STOCK for $10, and buying for $9.95 and you want to buy you can pl
288.
by ced 11y ago
Thank you for the detailed explanation, I get the point about liquidity and traditional market makers, that makes sense. But isn't the point about HFT (at least, the part that is described in the comic) being "insider information&
289.
by ced 11y ago
The bid ask spread is a measure of how far behind you are the moment you buy a stock. If the ask is $10 but the bid is $9.75 you've lost 0.25 per share the second you make a trade. That makes some sense, but who have you lost it to?
290.
by ced 11y ago
Have you tried Julia? Optimized Fortran (esp. on specialized hardware) would still beat it no question, but IMO Julia is a better value proposition for Fortran/C programmers than it is for Python/Matlab programmers at the moment.
291.
by ced 11y ago
How does that provide value? I have no personal experience in this, but I've heard that the order book is now very "shallow" due to HFT, such that it's hard to sell/buy a meaningful quantity of stocks at the market
292.
by ced 11y ago
it's still somewhat perplexing the extent to which vectorized code is necessarily slower The vectorized code typically allocates all kinds of intermediate results (more GC, more memory accesses). Apparently, turning it into loops is
293.
by ced 11y ago
I agree with your critiques of Python... Could you please post some example of code/operations which are very natural in R but unnatural in Python/Pandas? I'm curious to see what I'm missing out on.
294.
by ced 11y ago
Julia's for loops are comparable to C in performance, and its vectorized operations are comparable to Numpy/R, although some cases can be optimized using https://github.com/lindahua/Devectorize.jl (see the be
295.
by ced 11y ago
It's an interesting article. I haven't quite figured the difference to TD-learning, but this part in particular: in the spirit of temporal difference reinforcement learning, I hypothesize that certain neurons become critic neu
296.
by ced 11y ago
Ah, I took a similar path - applied Bayes. I was wondering where to go next for theory after Causality. It was such an interesting book.
297.
by ced 11y ago
Out of curiosity, where did you go (what did you read) after these two books?
298.
by ced 11y ago
Orbitz tackled a hard technical problem with Common Lisp. I don't know if it was the "secret weapon", but it must have been fairly critical. They needed efficiency. http://www.paulgraham.com/carl.html
299.
by ced 11y ago
That's a long list. Is it commonly believed that out-of-the-box Ubuntu is more secure than OSX?
300.
by ced 11y ago
There's also a Stan wrapper https://github.com/goedman/Stan.jl

More ›