VincentYu

• VincentYuMar 31, 2014, 3:44 AM

  4 points

  in reply to: John_Maxwell’s comment on: LessWrong Help Desk—free pa­per down­loads and more (2014)

  Here.

• VincentYuMar 30, 2014, 9:06 PM

  6 points

  in reply to: pianoforte611’s comment on: Open thread, 24-30 March 2014

  I’m currently learning about hypothesis testing in my statistics class. The idea is that you perform some test and you use the results of that test to calculate:

  P(data at least as extreme as your data | Null hypothesis)

  This is the p-value. If the p-value is below a certain threshold then you can reject the null hypothesis.

  This is correct.

  Put another way:

  P(data | hypothesis) = 1 - p-value

  and if 1 - p-value is high enough then you accept the hypothesis. (My use of “data” is handwaving and not quite correct but it doesn’t matter.)

  This is not correct. You seem to be under the impression that

  P(data | null hypothesis) + P(data | complement(null hypothesis)) = 1,

  but this is not true because

  1. complement(null hypothesis) may not have a well-defined distribution (frequentists might especially object to defining a prior here), and

  2. even if complement(null hypothesis) were well defined, the sum could fall anywhere in the closed interval [0, 2].

  More generally, most people (both frequentists and bayesians) would object to “accepting the hypothesis” based on rejecting the null, because rejecting the null means exactly what it says, and no more. You cannot conclude that an alternative hypothesis (such as the complement of the null) has higher likelihood or probability.

• VincentYuMar 15, 2014, 6:26 AM

  16 points

  in reply to: Alicorn’s comment on: What are you work­ing on? March /​ April 2014

  It’s intended for situations where Alice and Bob wish to communicate digitally with some degree of privacy and authenticity. At the moment, using software like GnuPG (which is a free and open source implementation of OpenPGP), Alice can encrypt her messages to Bob to provide privacy, and she can also digitally sign her messages to provide authenticity (i.e., a cryptographic proof that she wrote the message).

  Unfortunately, using such a system, Alice might worry that Bob could show others her messages and signatures, thus proving to others that she wrote and signed them. Such worries are not unusual if someone goes to the trouble of encrypting and signing their messages. For instance, if Alice is in a price negotiation with Bob, she might not want Bob to be able to prove to competitors that she offered a certain price.

  By using a ring signature instead of a standard signature, Alice can alleviate such worries. When she signs a message using a ring signature, Bob can be sure that Alice signed the message, but it only proves to everyone else that either Alice or Bob signed the message. Thus Bob is unable to prove to others what Alice said to him, because he could have forged the ring signature himself!

  Of course, Bob can still tell everyone what Alice said—using ring signatures instead of standard signatures only removes his ability to cryptographically prove to everyone what Alice said. This becomes a “he said she said” situation rather than a “she said this and here’s the proof” situation. From Alice’s point of view, the first situation is better than the second, so she would prefer to use ring signatures.

  More generally, I hope for ring signatures to become easy enough to use that a typical user can use it for normal emails. Right now, it is not advisable for the average user to sign their emails simply because it is often a bad idea to give recipients the ability to prove to others what they said.

• VincentYuMar 15, 2014, 4:46 AM

  19 points

  on: What are you work­ing on? March /​ April 2014

  Since December, I have been working on a proposed extension to OpenPGP to add support for some sort of designated verifier signature. Most of this work involved doing a literature review and carefully reading the OpenPGP specification to see what cryptographic schemes can be implemented successfully in OpenPGP. I settled on a ring signature scheme by Abe et al. (2004), and recently posted a proposal to the OpenPGP mailing list. Note that this is a proposal for a draft of a proposed standard, so things remain very flexible. This is the best time for anyone to comment on the proposal, before I write up an I-D.

  The following is a snippet of the description of ring signatures that I gave in my overview of the proposal:

  Ring signatures allow a signer to specify a set of possible signers without revealing who actually signed the message.

  The current OpenPGP specification lacks a straightforward way for Alice to send Bob an authenticated message that can only be verified by Bob. Right now, if such a signature is desired, the best that Alice can do is to send Bob a signed-then-encrypted message. Unfortunately, this means that Bob can show others the decrypted signature to prove to others that Alice signed the message. This ability of Bob to transfer Alice’s signature to others is often undesirable from Alice’s point of view.

  Ring signatures provide an easy way for Alice to authenticate a message to Bob without concomitantly giving Bob the ability to transfer her signature. Alice simply creates a ring signature that specifies Alice and Bob as the possible signers of a message. Upon receipt of the message and ring signature, Bob is assured that the message is signed by Alice, but he cannot prove to others that Alice signed the message because Bob could have created the ring signature himself.

  I am interested to hear from the perspective of end users. For instance:

  • Is there anything puzzling in my description of ring signatures?

  • Are there any particular features that you desire?

  Of course, I would also like to hear from anyone who has the time to go through the details of the proposal. Sections 3 and 4 contain the bulk of the actual cryptographic scheme, and it closely follows Abe et al.’s scheme with only small modifications to better fit OpenPGP. Some resources:

  • Ring signatures are interesting by themselves, and I think some Less Wrong readers might enjoy reading the following papers. The standard reference for ring signatures is Rivest et al. (2006); I would recommend reading this paper first if you plan on reading Abe et al.’s paper, because it contains a very clear exposition of the ideas behind ring signatures. A third paper that I recommend is Bresson et al. (2002); Sections 1–3 are particularly useful.

  • The current OpenPGP specification is given in RFC 4880. I don’t recommend reading it unless you have some burning desire to learn more about OpenPGP; it’s not particularly illuminating.

  In the best case scenario, ring signatures get deployed on GnuPG within a few years and Alice can create a ring-signed message for Bob simply by running: gpg2 --clearsign --ring-signature --recipient Bob

  (In the not-quite-worst case scenario, I write the I-D, everyone agrees it’s crap, and I vow never to touch math again.)

What are you work­ing on? March /​ April 2014

• VincentYuMar 6, 2014, 2:08 AM

  7 points

  in reply to: Pablo’s comment on: LessWrong Help Desk—free pa­per down­loads and more (2014)

  Here.

• VincentYuMar 4, 2014, 5:58 AM

  1 point

  in reply to: Pablo’s comment on: LessWrong Help Desk—free pa­per down­loads and more (2014)

  I’ve requested a scan from my university’s medical library.

• VincentYuFeb 26, 2014, 12:39 AM

  2 points

  in reply to: badger’s comment on: Open Thread Fe­bru­ary 25 - March 3

  There are some techniques that can be used with simulated annealing to deal with noise in the evaluation of the objective function. See Section 3 of Branke et al (2008) for a quick overview of proposed methods (they also propose new techniques in that paper). Most of these techniques come with the usual convergence guarantees that are associated with simulated annealing (but there are of course performance penalties in dealing with noise).

  What is the dimensionality of your parameter space? What do you know about the noise? (e.g., if you know that the noise is mostly homoscedastic or if you can parameterize it, then you can probably use this to push the performance of some of the simulated annealing algorithms.)

• VincentYuFeb 21, 2014, 10:25 AM

  10 points

  in reply to: gwern’s comment on: LessWrong Help Desk—free pa­per down­loads and more (2014)

  Here.

  apparently the fulltext is at http://​​sciencemag.lovefuck.me/​​content/​​343/​​6171/​​596.short—no, I don’t understand either why Science is now being hosted on lovefuck.me either. There are many things in this world I don’t understand.

  I took a look. That domain is acting as a 2-hop open web proxy. The first hop routes through a VPS in New York. The second hop routes through a (dedicated?) server in Montreal. The New York VPS is running nginx as a reverse proxy with no caching. The Montreal server is running Mr9.SM, which looks like an online fraud toolkit built on top of a web proxy back end. On the same server, there is an exposed MongoDB interface that is leaking data that should not be leaked.

  There are other domains that are also using these two servers as a 2-hop open web proxy (the domains and two servers are most likely managed by the same entity, since the setup requires coordination). A small sample (rot13ed to stop bots from picking them up):

  • nqbivan.pbz

  • pbaqbefubrf.pbz

  • qbjaybnqjvaqbjf8.zr

  • serrcfq.zr

  • ccggrzcyngrf.zr

  • jvaqbjfqbjaybnq.zr

  Strangely, I wasn’t able to figure out what fraud is being attempted. I expected to see cookie stuffing, but there was no sign of that—the only thing added to each page is a javascript StatCounter tracker. Phishing seems unlikely given the rather conspicuous domain names… but I suppose many users still fall for that. I don’t understand why a 2-hop design is used, especially since nginx is not being used to cache anything. If anyone figures out what’s going on, I would be very interested to hear about it.

• VincentYuFeb 21, 2014, 1:20 AM

  8 points

  on: A self-ex­per­i­ment in train­ing “notic­ing con­fu­sion”

  This is a well-written post! Upvoted.

  A nitpick:

  If I naïvely say that Week 1 establishes a true distribution for averaged weekly counts, then being more than 1σ above the mean for three weeks would have a probability of about p = (0.16)^3 = 0.0041 if that true count distribution remained constant.

  Unfortunately, this p-value is poorly calibrated because the sampling errors in the estimates of the weekly means and σ are non-negligible compared to the value of σ.* We can obtain an accurate p-value by simulation. Under the null hypothesis of no change in counting frequency, the count for each day follows a Poisson distribution with mean = 150 counts /​ 35 days (I got 150 from adding up all the counts in the plot; there is some sampling error in this estimate, but its effect on the estimated p-value is negligible). From simulating 10^5 samples, I found 8296 samples with week 3–5 means that are all greater than the sum of the week 1 mean and week 1 SD. This gives p = 0.083.

  An alternative (and standard) way to get a p-value here is to use Kendall tau as a test statistic, which gives a non-parametric rank-based test for monotone association. The single-tailed Kendall tau gives p = 0.076.

  * ETA: Let me add more explanation for any reader who is not sure what’s going on. The p-value in the post is (exactly) correct if the weekly mean and SD under the null hypothesis can be determined without any error. Unfortunately, we cannot do that—the best we can do is to estimate the weekly mean and SD using the week 1 mean and SD, so our estimates contain sampling errors. Often, we do not care about sampling errors when we are working with large samples because these errors are negligibly small compared to the SD. However, in this case, our sample has only n = 7, so sampling errors are non-negligible compared to the SD. This becomes a problem when we work with p-values because the null hypothesis is dependent on our estimates, but the errors in these estimates are not taken into consideration when we calculate the p-value. A common way to work around this is to use simulations, as I did. Alternatively, because our null hypothesis is rather simple, it might be feasible to use analytic methods to calculate a correct p-value.

  What links here?

  • A self-ex­per­i­ment in train­ing “notic­ing con­fu­sion” by whales (Feb 20, 2014, 1:55 AM; 65 points)

• VincentYuFeb 16, 2014, 7:58 AM

  5 points

  in reply to: RomeoStevens’s comment on: LessWrong help desk—free pa­per down­loads and more

  1. Here.

  2. Here.

• VincentYuFeb 16, 2014, 7:40 AM

  6 points

  in reply to: gwern’s comment on: LessWrong Help Desk—free pa­per down­loads and more (2014)

  1. Here.

  2. Here.

• VincentYuFeb 16, 2014, 7:40 AM

  7 points

  in reply to: gwern’s comment on: LessWrong Help Desk—free pa­per down­loads and more (2014)

  Here.

• VincentYuFeb 1, 2014, 5:34 PM

  0 points

  in reply to: Douglas_Knight’s comment on: LessWrong Help Desk—free pa­per down­loads and more (2014)

  Thanks!

• VincentYuFeb 1, 2014, 7:08 AM

  1 point

  in reply to: Douglas_Knight’s comment on: LessWrong Help Desk—free pa­per down­loads and more (2014)

  I am impressed; I had looked for the paper and failed to uncover the related patent. Could you share the exact string that you searched for (and the search engine you used) to discover the patent?

• VincentYuJan 28, 2014, 9:25 PM

  6 points

  in reply to: lukeprog’s comment on: Open thread, Jan­uary 25- Fe­bru­ary 1

  Wells’s book: Apocalypse when.

  I took a quick skim through the book. Your focused criticism of Wells’s book is somewhat unfair. The majority of the book (ch. 1–4) is about a survival analysis of doomsday risks. The scenario you quoted is in the last chapter (ch. 5), which looks like an afterthought to the main intent of the book (i.e., providing the survival analysis), and is prepended by the following disclaimer:

  This set serves as a foil to the balanced discussions by Rees, Leslie, Powell, and others. The choice of eight examples is purely arbitrary. Their purpose is not orderly coverage but merely examples that indicate a range of possibilities. The actual number of such complex unorthodox scenarios is virtually infinite, hence the high risk.

  I think it is fair to criticize the crackpot scenario that he gave as an example, but your criticism seems to suggest that his entire book is of the same crackpot nature, which it is not. It is unfortunate that PR articles and public attention focuses on the insubstantial parts of the book, but I am sure you know what that is like as the same occurs frequently to MIRI/​SIAI’s ideas.

  Orthogonal notes on the book’s content: Wells seems unaware of Bostrom’s work on observation selection effects, and it appears that he implicitly uses SSA. (I have not carefully read enough of his book to form an opinion on his analysis, nor do I currently know enough about survival analysis to know whether what he does is standard.)

• VincentYuJan 28, 2014, 6:21 PM

  21 points

  on: [Er­ror com­mu­ni­cat­ing with LW2 server]

  A fundamental issue here is that von Neumann–Morgenstern (VNM) utility functions (also called cardinal utility functions, as opposed to ordinal utility functions) are not comparable across entities; after all, they are only invariant up to positive affine transformations.

  This means that the relations in your post that involve more than one utility function are meaningless under the VNM framework. Contrary to popular misconception, the inequality

  u_v(v(1), b(0)) > u_b(v(0), b(1))

  tells us nothing about whether Veronica likes apple pies more than Betty does, and the equality

  u_b(v(1), b(0)) = u_v(v(0), b(1)) = 0

  tells us nothing about whether Betty and Veronica cares whether the other gets a pie.

  A quick way to see this formally is to note that you may transform one of the utility functions (by a positive affine transformation of the form ax + b, which leaves VNM utility functions invariant) and get any relation you want between the pair of utility functions at the specified points.

  A quick way to see this informally is to recall that only comparisons of differences in utility within a single entity are meaningful.

  Von Neumann and Morgenstern address some of these common misunderstandings in Theory of Games and Economic Behavior (3rd ed., p. 11; italics original, bold mine):

  A particularly striking expression of the popular misunderstanding about this pseudo-maximum problem [of utility maximization] is the famous statement according to which the purpose of social effort is the “greatest possible good for the greatest possible number.” A guiding principle cannot be formulated by the requirement of maximizing two (or more) functions at once.

  Such a principle, taken literally, is self-contradictory. (In general on function will have no maximum where the other function has one.) It is no better than saying, e.g., that a firm should obtain maximum prices at maximum turnover, or a maximum revenue at minimum outlay. If some order of importance of these principles or some weighted average is meant, this should be stated. However, in the situation of the participants in a social economy nothing of that sort is intended, but all maxima are desired at once—by various participants.

  One would be mistaken to believe that it can be obviated, like the difficulty in the Crusoe case mentioned in footnote 2 on p. 10, by a mere recourse to the devices of the theory of probability. Every participant can determine the variables which describe his own actions but not those of the others. Nevertheless those “alien” variables cannot, form his point of view, be described by statistical assumptions. This is because the others are guided, just as he himself, by rational principles—whatever that may mean—and no modus procedendi can be correct which does not attempt to understand those principles and the interactions of the conflicting interests of all participants.

  Sometimes some of these interests run more or less parallel—then we are nearer to a simple maximum problem. But they can just as well be opposed. The general theory must cover all these possibilities, all intermediary stages, and all their combinations.

  To directly address the issue of utility comparison across entities, refer to this footnote on p. 19:

  We have not obtained [from the von Neumann–Morgenstern axioms] any basis for a comparison, quantitatively or qualitatively, of the utilities of different individuals.

  I highly recommend reading the first sections of the book. Its copyright has expired and the Internet Archive has a scan of the book.

  A quick note for anyone confused over why the utility functions here are so much weaker than what they are used to seeing: You are probably used to seeing “utility” in discussions of utilitarianism, in which “utility” generally does not fall under the VNM framework (they are often intended to be much stronger than VNM utilities, so that they are no longer invariant under positive affine transformations and can be compared across entities—the trouble here is that there is no sensible formalization that captures these properties); that is, “utility” in utilitarianism suffers from a namespace collision with “utility” in economics and decision theory (“utility” in economics and decision theory also often refer to different things: ordinal utility is more common in the former whereas cardinal utility is more common in the latter).

• VincentYuJan 26, 2014, 10:32 PM

  12 points

  in reply to: jsalvatier’s comment on: LessWrong Help Desk—free pa­per down­loads and more (2014)

  Here.

  Does anyone know if Cochrane publishes the data they use in their meta analysis?

  I had a look. It turns out Cochrane does publish all their usable data, and they seem to be ungated! Here’s a link to the data for this meta-analysis. (The link to this data is provided in the gated HTML article, but there doesn’t seem to be a link from an ungated page, so I wonder if these data are supposed to be freely accessible… In any case, all their data are currently ungated and accessible by appending ‘/​downloadstats’ to the appropriate URL.)

• VincentYu 23 Jan 2014 6:14 UTC

  2 points

  in reply to: gwern’s comment on: LessWrong help desk—free pa­per down­loads and more

  Here.

• VincentYu 22 Jan 2014 19:31 UTC

  7 points

  in reply to: BraydenM’s comment on: LessWrong Help Desk—free pa­per down­loads and more (2014)

  Here.