12:42 AM
@Stephan On "mods commenting on unclear questions and requesting clarification. It seems like less than 10% of such questions are ever clarified" ... in spite of whuber stepping up to take a bullet there that problem is much more my fault than his, because I often don't follow them up. If I can't come back later (and really there's a bunch more stuff to deal with every time) I should lean toward closing on the spot.
Re failure to follow up or time-delayed closure ... I did ask for something related to that here -- something to make it easier to follow these up and close them (among other tasks). It was closed as a dupe of something I explicitly excluded from my question, but from the look of it would have suffered exactly the same fate (i.e. status-declined) if it had not.
1:42 AM
4 hours later…
5:22 AM
6:32 AM
7:09 AM
@Dawny33 I'd just post my comment as an answer: chat.stackexchange.com/transcript/message/30381000#30381000
Reinforcement learning is an area of machine learning inspired by behaviorist psychology, concerned with how software agents ought to take actions in an environment so as to maximize some notion of cumulative reward. The problem, due to its generality, is studied in many other disciplines, such as game theory, control theory, operations research, information theory, simulation-based optimization, multi-agent systems, swarm intelligence, statistics, and genetic algorithms. In the operations research and control literature, the field where reinforcement learning methods are studied is called ...
2 hours later…
9:35 AM
1 hour later…
10:47 AM
@amoeba I didn't read about how it worked, I simply assumed it would be very good at spotting patterns humans tend to come up with. So I used a sequence of numbers I didn't come up with that appears to be random but which I know (or rather, I know how it starts) and converted that into F's and D's (by the simple expedient of looking at whether it was even or not).
Of course to accept there was really something going on (like a claim of being able to "beat" the algorithm) I'd be using a substantially lower significance level than 5%
Usually I don't manage to recall quite so many digits. Most days I can get 50-odd before I struggle.
The other day I was talking to someone I had met a couple of times many years ago and he said "I always remember you because the second time we met you remembered my phone number from the first time we met. You said but it's easy, because it's divisible by 13."
He was more impressed that I could tell it was divisible by 13 than by the fact I could recall it a month later... which was actually the least tricky part, since you could see that at a glance
Often people who aren't used to some basic mental arithmetic can be inordinately impressed by simple tricks though
I've had students gape open mouthed when I do problems in my head quicker than they can type them into a calculator... even when there's only simple arithmetic
3 hours later…
2:03 PM
2 hours later…
3:49 PM
2 hours later…
5:50 PM
2 hours later…
7:51 PM
The grocery store by my apartment has the express lanes organized in a square, with two stations on each side, for a total of four stations.
And when there are two, they are always stationed on opposite sides of the square, forcing the customers to make two lines.
3 hours later…
11:14 PM
I was interested to watch the Mythbusters test out queueing theory -- they made a mock up grocery store/supermarket with volunteer customers and gave the customers varying lists of things to find and queue for and pay for (with fake money) and 10% (I think) of customers had something that would make checkout take longer. After running the system for a while they recorded how long people took to get from joining the queue to exiting the system
They did this for multiple servers with multiple lines and single lines. Their results were of course not sufficiently replicated but did suggest that the single line took longer on average. It made me wonder which assumption or assumptions failed that most contributed to what they saw.
« first day (2130 days earlier) ← previous day next day → last day (2881 days later) »