antonb2011

For chaos to manifest itself you need at least two things. First, the equations of the dynamical system have to be nonlinear. Second, the dimensionality of the systems has to be greater than 3. In quantum mechanics, all the equations are linear, so there is no place for chaos in a strict mathematical sense.

You suggest up/down sampling rare cases. Can you please elaborate on the standard approaches for this kind of problem? For both linear and nonlinear classifiers. Thank you.

"The Next Cryptocurrecy" is correct and is the phrase that tells you everything you need to know. The proponents of Bitcoin try to pass it off as being "As Good as Gold". However, unlike gold, whose supply is limited on Earth, and cannot be expanded without radical measures (asteroid mining or nuclear fusion), you can make as many new cryptocurrencies as you want. I might also have a couple of hash functions up my sleeve, whose solutions are difficult to compute, and so are "limited" in supply. Limited, that is, until someone comes up with next ByteCoin, CryptoCoin, ScamCoin etc.

Additionally, a shallow market is always a perfect pump-and-dump opportunity. Say you have a stock (or Bitcoin) that only has a volume of $100,000 a day. Then, if you're a small hedge fund and have about $20,000,000 devoted to small cap growth, or emerging market currencies, and your investment mandate allows you a temporary tactical deviation, you can buy the WHOLE supply of an asset for several WEEKS sending the price through the roof. When the crowd catches on and starts pumping money into your asset you exit the market and allow it to crash. Is it legal? No. Does it happen all the time? YES!

Not so. If it's a 4 digit combination, it's already n!, so 24 combinations, which is enough for a good system to lock up and automatically call the police. If it's more, then it's gonna grow as n!/k!l!m!... where, k, l, m... are the numbers of repeated digits, but still it's more than enough to know that someone's trying to brute force the system.

Well, viva human brain versions of Monte Carlo Search Tree algorithms. Your game was at least solvable deductively in practice. If you were talking about solving something like the game of Go, i.e. high branching factor and deep tree, then Monte Carlo algorithms would be your first guess instead of trying to do deductive reasoning. Upper Confidence bound on Trees (UCT) algorithm is a good one to read about and understand if you have a spare evening.

Hey, we should be thankful that primes have cardinality aleph_0, so we can have reasonable algos for exhaustive search at all! ;)

Well, this is cool, but it would be cooler if they wrote an open source cell tower for GSM phones running on Raspberry Pi.

I'd say also read "Seeing What's Next", the third book in the disruption series. There he deals with a lot of stuff like structuring your business to take advantage of Asymmetric Motivation and Asymmetric Skills as shield and sword of your business, as well as building an orthogonal Value Network to the existing solutions to avoid co-option by incumbents (or not doing so accidentally, if you are aiming to be acquired).

Whenever they tried giving me this test, my score just went off the scale. In the end, they had to assign me a new type: SEXY. ;)

Looks like the techniques of the world of High Frequency Trading are migrating into the wider universe. Except now the customers may benefit if the prices are lowered by someone who wouldn't sell at that price (by, say, showing that the product is in stock, but showing an error when you try to add it to your cart) in an attempt to make their competitors' algorithms follow the price drop, in order to deplete the stocks of goods in the process.

Wow! A HUGE market that escaped everybody’s attention! There are a few comments that come to mind:

First, this Modj.io company provides free testing ground for car companies to see if there is really a wide demand for such feedback mechanisms in public. At the first sight, I would say that there is, as they outlined in their sample use cases in the video.

Second, the only chance that Modj.io has for survival is a quick ramp-up of user base, because this technology is a perfect example of what car companies will try to use for customer lock-in. Kind of like iTunes for your car, storing driving and tracking data in the cloud. I foresee that every car company will try to develop their own platform, SDK, and an app, with unique synergies coming from having access to their cars’ deep engineering knowledge. Additionally, this will give car companies real-time feedback about the performance of their cars down to the last part that failed. This will allow them to know things like that the left windshield wiper motor batch that fails more often that average was assembled on a Friday night by John Doe. That’s scary level of detail and feedback.

Third, with such level of granularity, insurance companies will have a whole new set of market segmentation metrics, and I wouldn’t be surprised to see attempts in horizontal integration of car companies with such feedback information with insurance providers.

Forth, when each car company starts promoting their own SDK, we will see an aggregator-type company that will produce an abstraction SDK that allows you to write an app once to work on all cars.

Just my $0.02. I’d love to hear your projections as well, or comments on something I missed.

Don't worry about it, man. He can patent it all he wants, but thing is not gonna fly. It's called a "tie-in sale", and the United States Anti-Trust law prohibits it. McDonald's tried this trick a long time ago with their franchise owners in order to better account for how much they sold, by forcing them to buy paper cups and other supplies directly from Corporate (i.e. prohibiting third party suppliers). I think it got really high in courts, but eventually the franchise owners won.

That's what you think. A good friend of mine is 6'2 and used to be a part time male fitness model. He's a software engineer at a large bank and all the girls hit on him and never believe him when he says he's a software engineer.

Good read. This coincides with my own impression after reading HN for a while. I suppose the problem of "Feature, Not a Company" sets in for any maturing market. But aside from "plumbing" start-ups discussed on HN (like Heroku, Phpfog or Parse), I rarely see anything that would be viable as a long term stand-alone business, as many of our business-side colleagues would also assert via "Porter's Five Forces"-type analysis. So I guess now it's all about making a quick lottery ticket type start-up and trying to get acquired. That's why you need VCs, 'cause they can negotiate a much higher price for your little company that you can.