kylebrown

14 ms·

61.
by kylebrown 12y ago
Now they just need to release the datasets as well. The datasets are far more important than the algos.
62.
by kylebrown 12y ago
There were several functional alpha clients even before the crowdsale. PoS vs PoW is somewhat beside the point anyway, as the codebase is still using SHA3 PoW: http://github.com/ethereum
63.
by kylebrown 12y ago
Clinical medicine is a better analogy than technical rollouts. But you're right, modern development economics argues that social fixes (policies) should be empirically driven using evidence-based randomized control trials. Esther Duflo
64.
by kylebrown 12y ago
What are the better alternatives for allocating resources? Any distribution mechanism will need a standard metric for measuring and dividing piles of resources. Money is that metric. I do agree with you about work though. Assuming a future
65.
by kylebrown 12y ago
The roadmap is 15 years, but the project was funded in October 2008. With six years of work behind them, how far down the 15-year roadmap at present?
66.
by kylebrown 12y ago
I have no idea what your then-girlfriend meant by one of those guys. Or whether the Bronx hood was full of obsessed guys in the '70s?? I am genuinely curious, though.
67.
by kylebrown 12y ago
There are benefits to fasting for 10 hours: http://www.sciencealert.com/new-research-suggests-you-make-b... The research was also summarized as "always gamble on an empty stomach."
68.
by kylebrown 12y ago
What are the physics behind the 20mb/s bandwidth limit over radio? He said more than that isn't "feasible", because of power requirements or why?
69.
by kylebrown 12y ago
How do you know that humans don't use intelligent bruteforce approximations to win?
70.
by kylebrown 12y ago
If you input a score ("{-1,0,1} on final positions") its effectively a label, that makes the training supervised rather than unsupervised. See [1] for good reasons to be skeptical of unsupervised learning in general. See [2] for a
71.
by kylebrown 12y ago
Its trivially easy to write some formulas down on paper ;) I was looking up AIXI the other day, but couldn't figure out what the acronym stands for.
72.
by kylebrown 12y ago
The puzzle is not figuring out how to log in. You get an invite, then you get the challenges: http://i.imgur.com/xtCdf94.png They are story problems. The first one is cycle detection for a singly-linked list. After solving
73.
by kylebrown 12y ago
I got an invitation a couple days ago when I searched "mutex lock". Not sure whether search history makes a difference, but it worked for the gf too a couple minutes later: http://i.imgur.com/l4rGps5.png
74.
by kylebrown 12y ago
Here's deep Q-learning implemented in a javascript demo: http://cs.stanford.edu/people/karpathy/convnetjs/demo/rldemo...
75.
by kylebrown 12y ago
The online poker sites operated offshore, through high-risk credit card processors. Fantasy sports markets, otoh, have Yahoo and CBS as operators (both take ~30% rake). FanDuel is processing a few hundred million through PayPal. Presumably,
76.
by kylebrown 12y ago
I'm actually surprised you didn't overfit the data until the backtests showed you had a profitable algorithm. I bet most people would've done that.. I'm curious, what kind of algo were you training?
77.
by kylebrown 12y ago
> The house is taking 8 - 10% rake meaning you have to have a 60% winning percentage just to break even. That's only correct in a constant sized market. In a growing market with lots of new money coming in, its much easier to win. W
78.
by kylebrown 12y ago
I had fun trying my hand at generating NBA lineups earlier this year.[1] I'd love to give it another go (my email is in my profile). 1. https://news.ycombinator.com/item?id=7272365
79.
by kylebrown 12y ago
Another option is to match orders with frequent batch auctions (instead of continuous double).[1] 1. http://faculty.chicagobooth.edu/eric.budish/research/HFT-Fre...
80.
by kylebrown 12y ago
Even mentioned transducers, cool! But, I was hoping for a treatment of CSP and react/flux. jlongster teased us about that here[1]. Maybe the next post in the series? (plz?) 1. https://justin.harmonize.fm/development&#x2
81.
by kylebrown 12y ago
I looked around the site, but couldn't find anything about turning ball-point sketches into 3d models.
82.
by kylebrown 12y ago
On this, Alan Watts offers some sage advice: "If you get the message, hang up the phone."
83.
by kylebrown 12y ago
Hal Abelson disagrees with you. "Computer Science is a terrible name for this business. First of all, its not a science. It might engineering, or it might be art. But we'll actually see that computer -- so-called science -- actual
84.
by kylebrown 12y ago
Wow, this is superb educational content.[1] Is the author anonymous? 1. also https://curiosity-driven.org/low-level-bitcoin
85.
by kylebrown 12y ago
Well, the one thing Stellar has going for it is that Stripe is on board. Though afaict Stripe is not yet integrated as a Stellar gateway. The only thing keeping Ripple alive is the fact that Bitstamp is a Ripple gateway.
86.
by kylebrown 12y ago
Yup, obviously a fork of Ripple. While at Ripple, Jed acqui-hired his gf Joyce (SimpleHoney). Then MtGox blew up (Jed still had his profit-sharing agreement with MtGox from his sale to Mark Karpeles). Then Jed announced he was going to dump
87.
by kylebrown 12y ago
Hi Joyce, would you ask your bf Jed where his MtGox profits are? Its known that Jed sold MtGox to Mark Karpeles, and had access to a privileged account which allowed him to audit the total trading volume (and collect his agreed profit share
88.
by kylebrown 12y ago
These things will work, they just won't be good for "store of value" like bitcoin. Supply of disk space, bandwidth etc keeps growing, but the demand for those resources grows as well. Hard to predict the ratio of supply to de
89.
by kylebrown 12y ago
The idea is that a lot of devices wouldn't need a two-way connection, rather they could confirm payment with only a downlink.
90.
by kylebrown 12y ago
Reversing an unconfirmed (0-confirmations) transaction is easy to do, with a supposedly 10-20% success rate, and doesn't depend at all on hashing power. Just broadcast a 2nd conflicting tx that pays a higher fee (most nodes won't

More ›