clickok

10 ms·

31.
by clickok 8y ago
Yes. I don't know what your thresholds for impressive might be, or what your basis of comparison is, but pure control theory seems to be at a bit of a dead end, whereas reinforcement learning allows for greater flexibility and robustne
32.
by clickok 8y ago
In imperfect information games (like SC2) the outcome prediction implicitly takes into account the unknown. Given what it has observed and what it has not observed, it is essentially comparing the present state to similar situations from
33.
by clickok 8y ago
It does, insofar as you can express anything meaningful in an inconsistent system. A formal system being inconsistent implies being able to prove some statement A , and also its converse ~A . If both A and ~A are true, then we can pro
34.
by clickok 8y ago
Assuming that insurance is already priced as accurately as it can be according to the available data, then discovering that wealthy people tend to evacuate disaster areas while the poorer residents remain would suggest that you charge great
35.
by clickok 8y ago
In the article it mentions that the data comes from apps on the phone, so insofar as the user has clicked through the EULA, Privacy Policy, and permissions screens[0], it's voluntary in theory but obviously not in practice. The interes
36.
by clickok 8y ago
Read the list of individuals to whom this "sophisticated" personality clustering technique was sent. You've got people like A.V. Aho, R.W. Hamming, J.B. Kruskal, J.R. Pierce, K. L. Thompson, J.W. Tukey (and probably more who
37.
by clickok 8y ago
So how long would you expect books bound this way to last? I took some pictures of the ones I've made for reference: https://imgur.com/a/60zP8vn . So far, none of them have come apart but I'm pretty careful
38.
by clickok 8y ago
I took some pictures to demonstrate the final product[0]. I am not an expert bookbinder, but it works well enough and is cheap and easy to do by hand. This is the result of tinkering; presumably someone who knows what they're doing cou
39.
by clickok 8y ago
The problem with most recent books is that they use a glue-based binding, but the glue seeps into the spine far enough that it makes laying them flat difficult. However, if you bind it yourself, you can get the desired behavior pretty easil
40.
by clickok 8y ago
I am not sure that he's the first to note that the brain is a predictive machine, or to notice that models of state or the environment can be formulated entirely in terms of predictions[0]. Personally, I find that idea very convincing,
41.
by clickok 8y ago
Or just a private firm[0], since models need training data and Facebook has a truly staggering amount, all pre-tagged. There is also a small amount of irony in the fact that Facebook's initial incarnations were pre-seeded with scraped
42.
by clickok 8y ago
Originally I found the Magic Leap extremely exciting, because in contrast to phones or tablets, whose touchscreens limit what sort of applications are feasible, AR could allow for arbitrary interfaces. These interfaces could be present alo
43.
by clickok 8y ago
Yes, you could have different possible production rules for a given symbol and sample from them according to some distribution. You could also just randomly perturb/fuzz a given iteration of the system. A deterministic system is nice,
44.
by clickok 8y ago
That might defeat NLP-based analysis, but the use of obfuscation and the methods employed still tells you something about potential authors. If someone in an organization writes an essay disparaging that organization, but does so in an unna
45.
by clickok 8y ago
There are a couple of places where RL/Evolutionary Methods/Something Else are jockeying for the top of the leaderboards, but usually if one technique succeeds over the others it just means that someone's put a lot of work int
46.
by clickok 8y ago
Pretty cool, this is actually a great reference for a lot of things. Even if you're familiar with RL, you might be reminded of something or learn something new. Sutton and Barto's book is also good if you want to do more than dip
47.
by clickok 8y ago
I skimmed through this and have already found a bunch of interesting sections, but there's also a ton of background information on topics related to bandit algorithms. The authors say that this is the first draft of the book submitted
48.
by clickok 8y ago
It doesn't have to be literally false to misinform, and informing the public is (in my opinion) the main point of journalism. How erroneous/slanted does it have to be before we can assert fakeness? I think it's fair to say t
49.
by clickok 8y ago
I am not so sure; there's rarely a strong incentive to misrepresent a paper on e.g. number theory, but that is not the case when it comes to issues in politics or business. We already see this to some extent with Wikipedia-- the pages
50.
by clickok 8y ago
It is, and has been for decades. Usually, when I read articles on a topic which I have researched, I find the reporting tends to be "wrong" in some sense. Charitably this can be blamed on time or space constraints: the articles
51.
by clickok 8y ago
It sounds like their sudden viral popularity created too many obligations to fulfill, but not enough to scale up sustainably. Truly victims of their own success. I know that stretch goals are popular but there's something to be said fo
52.
by clickok 8y ago
I would not dive into GANs right off the bat. Ng's course is a good start, maybe also buy or borrow a copy of Russel and Norvig's "Artificial Intelligence: A Modern Approach"[0]. If you decide to look into deep learning,
53.
by clickok 8y ago
It's interesting that so much neural net weirdness emerges from exploiting errors in physics simulators or floating point math. I am now expecting the next generation of perpetual motion machines to include AI to try to take advantage
54.
by clickok 8y ago
I'm glad to hear that. The main point I try to get across to people regarding Bellman equations is that they are very special-- these sorts of recursive equations allow us to express the value of an observation without knowing the past
55.
by clickok 8y ago
Yeah, so here we can "solve" the problem without using a POMDP by really expanding what constitutes the state space. Initially, your state is your hand plus the initial rules (who goes first, which suit is trump, etc depending on
56.
by clickok 8y ago
Consider reading "Reinforcement Learning, an Introduction" by Rich Sutton[0]. It is very accessible (and probably the most used textbook in the field), and goes into detail about the connection between RL and MDPs. If you want a h
57.
by clickok 9y ago
Yes. In particular, it's possible to learn the variance of the return using TD-methods with the same computational complexity as learning the expected value (the value function). See [0] for how to do it via the squared TD-error, or [1
58.
by clickok 9y ago
I can't comment on #4 but with regards to 1-2 we can make an argument that high dimensional space is generally friendly towards stochastic gradient descent. Consider a neural net that produces a single output given `N` inputs– it'
59.
by clickok 9y ago
Might it have been Lyrebird[0]? They're (as far as I can tell) they have the best text-to-speech that you can actually use. It's kinda annoying when Google makes these announcements about their advances in TTS only to reveal that
60.
by clickok 9y ago
According to: https://www.rmv.se/aktuellt/det-visar-tre-manader-av-medicin... I don't speak Swedish, so I had to translate it (the site I found the link from[0] after Googling seems like it has an axe to grind). T

More ›