Below are the 20 most recent journal entries recorded in Mark Gritter's LiveJournal:

[ << Previous 20 ]

Thursday, November 26th, 2009
9:02 pm	17-card Chinese Poker hands Is there a better way to set either of the following hands? 5c4c3c2cAc / 5h6d8hTdQs / 4h5d7s8c / JhJdKh The straight flush seems so appealing, but if there's no royalty for it is it worth trying to improve the Badugi and 2-7 hands? (But then it looks difficult to put JJ in front too.) 8h8s8cJdJh / 9d7d4s3c2s / 9s7c6h2d / AcAhQs Could play #2 for the 2-7 hand by abandoning the 9-high badugi, which isn't that great anyway. Rules: 5-card high hand, 5-card 2-7 hand, 4-card Badugi hand, 3-card high hand which must be less strong than the back hand. No royalties or naturals. If score matters, figure 4 points for winning 3 (1 point per hand + 2 points for overall winner) and 6 points for a sweep, with no tiebreaker. (What is the typical scoring? This seems to not reward a sweep enough.) (3 Comments |Comment on this)
1:08 am	Google Wave robots Tonight I spent some time playing around with the Google Wave Robots API, which lets external applications react to Google Wave events. This also required me to get familiar with the Google Apps service, which I hadn't previously used. Both seem relatively easy to use, and it's very nice to have a free 'server' for trying out cloud applications. The docs on the Robots Python client library would have been more helpful if they'd said "all the objects you care about live in the model module", and I'm still not entirely sure where I'm supposed to be looking to find out what information gets sent in which event. I am thinking of writing a Wave-to-LJ bridge; the pieces should be relatively easy to put together. Unfortunately, while looking at the LJ side for the first time in a few years (**), I re-discovered that its XML-RPC API doesn't permit addition of comments. So now I am not at all sure what such a bridge should look like. A wave plugin that makes me happy: a gadget which renders LaTeX in comments. I've wanted this in email and IM for quite some time. Not that I send a *lot* of equations around, and usually to people who could just parse the LaTeX syntax. (MathML is a sick joke. Better than embedded images, but still sick.) (**) since my Advogato-to-LJ bridge, come to think of it... (2 Comments |Comment on this)
Monday, November 23rd, 2009
1:18 am	License Plate Vocabulary mrissa introduced me to a simple license-plate game: given a three-letter license plate, is there a word which uses those three letters in that order? I've been curious what percentage of possible three-letter combinations have no solution. (A somewhat harder question is what percentage of valid license plate combinations have no solution, as the selection is usually not truly random.) Using SOWPODS as my word list, I discovered tonight that 1385 three-letter groups have no words. That is approximately 7.9% of the 17,576 possibilities. ( Here's the complete list. ) The python script I wrote scans the word list in a single pass, and keeps a tree of three-letter groups not yet found. My hope was that this would allow us to prune the generation of combination from words--- if we already know that all ABx are present we don't need to work through the remaining six letters in ABSOLUTE (similarly for ASx, ALx, etc.) But I'm not sure how big of a difference this actually makes. Guess I should have kept stats. ;) (9 Comments |Comment on this)
Saturday, November 21st, 2009
6:10 pm	MinneBar 2009 I had fun at MinneBar, despite what the following may sound like. I think the organizers did a great job finding space (the Best Buy headquarters) and getting everything lined up. Next year, I am totally signing up to lead a session or give a talk, because I was under-impressed with a lot of the ones I attended. So, hey, there's one sure solution to "I could do better than that." After I got home I tried to read twitters from others who attended. What an annoying process. Is there any way to filter out, say, retweets? Intellectual Property Tips for Entrepreneurs: Little preparation, little content. Lots of patent talk, no actual lawyers. Indie Video Game Development: Could have been very interesting, but too much time spent 'going around the room', doesn't scale to 500-person (un)conferences. Some grousing about the difficulty of getting 3D engines going, little talk about actual *game* design. Nathan Yourchuck and I scheduled our own impromptu session on "Economic Models for MMPOGs" but got no takers. Algorithms and Refutations: an excellent talk! (Except for tendency to leave white slides in between content... why?) Argues that bugs aren't "shameful", but inevitable due to fuzziness of concepts, just as counterexamples and concept-stretching lead to better proofs and concept-formation. Doesn't really seem to connect with studies of what causes *actual* bugs, though. How does this theory explain popular bugs such as off-by-one errors? Physical APIs: caught the end of this one, Jeremy Lizakowski showing off electronic parts. Fun, sort of. Discovering Scala: disastrous. Do not put up the project web page as your introduction and then ask for questions. A presentation on a new technology should sell that technology first, otherwise why do I care? I didn't care, so I left. Bootstrapping your Tech Start-up: Fun, not all that useful. Fighting broke out over his classification of marketing as a "black hole" for $$. Too much FUD over VC funding. Bad stats: "625,000 companies started in 2009, only 618 had venture funding". How many are restaurants, retail stores, caterers, service companies, etc., vs tech companies? Of tech companies that had a successful exit, how many were venture-funded? On further reflection, what annoys me most is that 90% of the talk was on saving money but only 10% was on earning money. Entrepreneurs don't need to be told that office space costs money. How to convince somebody to give you money for an incomplete product is something entirely different. Minnesota Lean Startup Group: also a disaster. Trying to launch a discussion group, but didn't do a good job presenting what he wants to discuss. I left. Understanding Monads: Interesting talk, trying to demonstrate monads implemented in JavaScript. Unfortunately, his examples were not very compelling--- seemed to be introducing much more complexity than the translation to monads was removing. (2 Comments |Comment on this)
Friday, November 20th, 2009
8:54 pm	Coolest Thing Ever OTCAMP: A Life pattern which simulates any outer-totalistic cellular automata, including Life itself. Each simulated cell is 2048x2048 cells large. With Golly you can run fairly large-sized patterns... but, it's only feasible to run with the HashLife algorithm, which cheats by memoizing the evolution of Life patterns rather than simulating every cell. Golly includes a few patterns using OTCAMP as samples. (I totally need to upload my PokerStars icon here, if I can remember where I put it...) (Comment on this)
Wednesday, November 18th, 2009
10:58 pm	*Who's* got unrealistic expectations, again? OkCupid has another great entry examining how physical attractiveness affects your chance of getting a response to a message. Unsurprisingly, people with more highly-rated pictures get better response rates (except in the case where a highly-rated male contacts a low-rated female, or vice versa.) But the really surprising point to me in this entry was the distribution of male and female ratings: These could not be more different. Males rated females' pictures on a normal-ish curve. (Incidentally, I didn't find the examples of "most attractive" women all that interesting--- either the two there now, or the previous example in an accidental early version of the post.) This is not to say men don't have unrealistic expectations; when men send messages on OkCupid, it is pretty dramatically skewed to women with high attractiveness ratings. But, this is more or less what you'd expect from an "objective" measure: as many attractive as unattractive, with most women in the middle. Females, on the other hand, are convinced that fully 80% of men (on the dating site, anyway) are less attractive than average. Wow. Theories? Men are less interested in personal grooming? Women are more attracted to non-visual stimuli? Women are just more picky? Women's normal curve is centered around 1 star (that is, they would have given plenty of negative stars if available)? P.S. OkCupid loses points for poor chart design, though. How big is each unit on the horizontal axis supposed to be? (8 Comments |Comment on this)
Monday, November 16th, 2009
11:11 pm	Classifying 2-D Cellular Automata David Eppstein (who is not one of the missing Eppes siblings) announced a new paper on classifying 2-D cellular automata, entitled "Growth and Decay in Life-Like Cellular Automata". (announcement). This is a subject dear (if not near) to my heart. At Gustavus, I took a seminar on "Complexity" from Nobel laureate Philip Anderson, who was there as a guest for a semester. My seminar project was on characterizing 2-D cellular automata--- specifically, looking at whether the simple lambda metric devised by Christopher Langton was a good predictor of behavior. (Lambda measures what fraction of output states are quiescent, and it appears to be a good fit for 1-D automata.) My conclusion was that it was not, because I found both examples of complex behavior far from the 'critical value' and many examples of chaotic behavior close to the critical value. (Digression: This was not a project that Dr. Anderson was particularly interested in. He suggested a much more difficult topic that I still have little idea how to tackle--- basically, how to characterize the search space of "realistic" problems. Computational complexity is full of examples that cause search algorithms to produce worst-case behavior. Can we make some sort of interesting statement about which properties of the search space produce "good" behavior? Obviously trivial problems can be solved trivially...) Eppstein argues that Wolfram's four categories of CA behavior are insufficient because they ask about behavior starting from a random state. But interesting examples of Life behavior are highly nonrandom starting states--- random Life patterns tend to settle into still lifes and oscillators, so why is it not "Category II"? Eppstein proposes a different categorization based on just two axes: does the CA have patterns which grow to escape any finite bounding box, and does it have patterns which "die out" completely? An advantage of these criteria is that we can actually reason about them based on the rule set, instead of making qualitative judgements by observing behavior. Some of Eppstein's earlier work was on searching for spaceships in semi-totalistic CAs like Life --- it's in More Games of No Chance, which I would strongly encourage math (and game) geeks to pick up should they have the chance. (Volume 3 in the series is out and on my Christmas list...) (Comment on this)
Sunday, November 15th, 2009
4:43 pm	Bah Game computer is still dead. The new power supply checked out fine; after I installed it I got fans and LEDs and everything else expected, but didn't hook it up to the TV to make sure it booted. Well, I did that today and it powered itself off after just a few seconds. It was willing to make one or two more attempts at booting but now it's back in a bricked state. (But with +12V everywhere there should be!) I hate computers. (Comment on this)
Wednesday, November 11th, 2009
9:46 pm	Not Entirely Given Up On Instant Gratification While I didn't find anything to buy at Fry's yesterday, I did visit BookBuyers this evening and found Saunders MacLane and Garrett Birkhoff's Algebra and could not resist. (Comment on this)
11:42 am	Keyboard! I gave up on instant gratification and ordered this Adesso split keyboard (with touchpad) as a replacement for my dying Logitech. I also got just as good a price at NewEgg for my replacement hard drives as I saw at Fry's. ($55 for 500GB 3.5" SATA.) (Comment on this)
Saturday, November 7th, 2009
8:19 pm	Half-Dead Power Supply? My gaming computer bricked: no lights, no fans. Naturally that led me to be suspicious of the power supply. The little Shuttle boxes are packed quite tightly so it was not entirely straightforward to get the power supply extricated. But when I did, I noticed that it made a little sparking/clicking noise when the power was applied. Also, all the +12V outputs seemed to be between +6V and +8V. So, diagnosis confirmed. Fortunately I was able to just look up the power supply model number on Google and find many people willing to sell me a replacement. I was also able to confirm that falling voltage is a common symptom of failing power supplies. But, I'm not clear from an electronics perspective why that should be. I've put together a switching power supply or two from components, but just DC-to-DC. My gut feel is that the order the components would fail in is: (1) fan, (2) "switch", (3) rectifier, (4) capacitor, (5) inductor. But I don't quite see how a switch failure would still provide DC current, just not enough. Maybe it's the capacitors that tend to go first, and thus they don't hold enough charge to reach the desired voltage. I wonder if anybody has done a study... maybe I should ask Patrick, our platform guru. Or my father, an electrical engineer. ETA: I found this rather dense and jargonish PDF presentation from IBM. Their failure percentages are: ~30-35% MOSFET (amplifier?), 10-15% Choke (inductor), 10-15% Driver (switch?), 0-5% Ceramic (capacitor), 0-5% Poly (poly-propylene capactior), < 2% Fan. So I guess I was totally off base on fans and inductors. (1 Comment |Comment on this)
Wednesday, November 4th, 2009
4:03 pm	Saving is for Suckers, too Wells Fargo is paying 0.05% on savings accounts. 5 cents per $100. I did let them talk me into an automatic savings account which pays 3% on the first $500 and then 3% on the first 13 months of deposits, which is way better than the other alternatives I'd seen. And now I get 0.03% on checking, which is admittedly better than nothing. Still... pathetic. (4 Comments |Comment on this)
12:27 am	MinneBar 2009 MinneBar is November 21st--- coming up fast! Unfortunately they have already run out of room and aren't accepting any more free registrations. It's being held at the Best Buy headquarters this year. I am thinking of participating more actively this year and either giving a presentation or trying to get on a panel. One topic already posted which interests me is "Algorithms and Refutations", trying to apply Imre Lakatos' insights on mathematics to software development. I think a "Bay Area startups vs. Midwest startups" would be an interesting panel, but I don't really have the contacts to put this together. I'm also considering a presentation/demo of Stasis, a library we're considering using at Tintri--- that would be a more self-contained session. (3 Comments |Comment on this)
Tuesday, November 3rd, 2009
12:53 am	Saving Grace Marissa watched the pilot of "Saving Grace" as potential workout fodder but was nauseated by the presence of chewing tobacco. (Decomposed corpses in "Bones" are OK. I can kind of understand this.) I was nauseated by Earl the angel for other reasons. But, let's take him as given. What is the appropriate response to somebody asking about the question of evil? If for whatever reason Earl doesn't want to give a straight answer--- he issues some mealy-mouthed statement about that not being what he's there for or some such--- a better-written angel would suggest that Grace should do her homework first! This seems such a consistent trope in "spiritual" shows; they treat problems of theology as if they are fresh and mysterious "big questions". Just like science fiction shows... and science journalism. But, humans have actually put a lot of thought into these questions over the years. Obviously Grace is not asking about evil because she wants to know, but because she wants an excuse not to know or believe. But I still think a reasonable response from Earl is to point her at Theodicy on Wikipedia and have her come back when she's figured out what she doesn't like of the (many) answers already devised. (5 Comments |Comment on this)
12:12 am	Customer Loyalty is for Suckers I'm celebrating my new Northwest Gold Elite/Delta Gold Medallion frequent-flyer status by instead taking United Airlines to SFO on my next trip. ($200 cheaper to get home on Friday instead of Saturday? Sold.) The return trip is through Denver. I've had good luck flying Frontier through Denver--- no missed connections in several years of travel. Rather the opposite with United, unfortunately. I'll be out in California for three more weeks before the end of the year, including the first two weeks of December. I'm trying to figure out if I'm better off booking a round-trip ticket to come home on the weekend, booking two Monday-Friday tickets, or if I just want to stay the weekend. (Comment on this)
Monday, November 2nd, 2009
2:19 pm	Sold Gold, Baby I suck at math or something, because I didn't think I'd make it to Gold frequent flier status on Northwest/Delta, and yet the last trip put me over 50K miles. 10K free qualification miles in July + 10 round trips @ 3178 miles = 42,000 miles this year, where's the extra 8K come from? Omaha + Grand Rapids trips = 3 segments, 500 miles each = 1500. Aha! I got awarded elite qualification bonuses (for business-class travel or seasonal specials or whatever) totaling about 7,000 miles. (2 Comments |Comment on this)
12:13 am	Quantum Search Engines "Privacy and the Quantum Internet" (in the October 2009 Scientific American) seems like a very cute and completely useless idea. The authors' proposal is to implement an eavesdropping-free "quantum private search" which makes use of quantum entanglement to allow the sender to verify that her query has not been observed. (See, for example, their security analysis.) In fact, a group at the University of Rome has demonstrated such a system built on "quantum RAM" which performs two index lookups simultaneously, given two queries in superposition. This works great as long as you are doing retrieval of an entry in a fixed database. Forget about asking queries that aren't precomputed, though. Admittedly, most two-word queries could be encoded in the 30-bit table they suggest. A larger problem, however, is that their security analysis depends upon Bob replying to Alice's query. A future quantum internet which does not drop any packets would be radical indeed! It seems completely infeasible to imagine a communication system with no error or noise, and thus the possibility of lost or damaged queries must be part of the protocol. But if Alice is willing to resend a query without a response from Bob, then Bob can take all the measurements he wants of the initial query, and simply fail to respond. In fact Bob (or Larry or Sergey) does not have to collect queries at close to the rate of natural error. Bob can collect several minutes worth of queries, measure them, and simply claim to have experienced a power failure or other data center fault. Admittedly Bob will have worse end-user performance than competitors who do not cheat... A review of their paper, mentioning an additional restriction: Alice must send her queries one at a time. The requirement of noiseless communication is brought up incidentally as a restriction, but I think it's closer to a fatal flaw. (Comment on this)
Saturday, October 31st, 2009
4:29 pm	Integer sequence puzzle The week a co-worker brought in a puzzle her child had been assigned in math class. The beginning and end of the sequence were given as 4 and 134. I offhandedly remarked that I would just cheat using the Online Encyclopedia of Integer Sequences. By now, of course, I've forgotten how many entries there were in the sequence! There were at least 4. Can I still use the OEIS to identify it? 4, _, _, _, _, 134: 4, 6, 26, 34, 86, 134, ... are "interprimes which are of the form 2*prime". (A075277) An interprime is the mean of two consecutive primes. A047732 and A056309 are good candidates too but they have prefixes. 4, _, _, _, _, _, 134: Some interesting candidates but none without a prefix. 4, _, _, _, _, _, _, 134: 4, 26, 28, 52, 76, 98, 124, 134, ... are "middle side of the first primitive d-arithmetic triangle" (A047732) which I am unable to decipher. I particularly like the Moebius Transform of Fibonacci numbers 1, 0, 1, 2, 4, 6, 12, 18, 32, 50, 88, 134. A007436. 4, _, _, _, _, _, _, _, 134: 4, 7, 40, 49, 59, 70, 85, 125, 134, ...: numbers n such that n^12 reversed is prime! Obviously A059213 is what was desired, it's so obvious! 4, 9, 15, 26, 39, 58, 85, 111, 134, ... are "semiprimes with triangular indices" (A122964) which is also cool. 4, _, _, _, _, _, _, _, _, 134: 4, 14, 34, 38, 40, 56, 80, 110, 128, 134, ...(A032645) are numbers n such that n concatenated with n+5 is a lucky number. 4, 5, 7, 8, 9, 13, 15, 52, 82, 134, ... (A119642) are indices of prime numbers of trees with n unlabeled nodes, which is neat but not a "basic" sequence. ... and we seem to be getting more candidates with bigger gaps. Alas, none of these look like an appropriate grade-school "trick" question. (8 Comments |Comment on this)
12:53 am	We Don't Live in Saint Paul Dammit I *really* wish that our mortgage company understood we live in Eagan and not St. Paul, no matter what the USPS database says about our ZIP code. Nearly everybody else manages to get this right. But even when I called them up and tried to get them to fix it a few years back, the customer service rep said there was basically nothing he could do--- their internal systems just wouldn't accept Eagan as a valid address. (8 Comments |Comment on this)
Friday, October 23rd, 2009
12:53 pm	Reality >>> Theory Bianca Schroeder's talk on DRAM errors is worth watching. (Though it did remind me how annoying EE380 sessions could be--- it's generally not students asking the same question over and over, it's faculty or random people off the street.) Old theory: Increased levels of heat cause errors, soft errors dominate and are randomly distributed. (Lab testing involves pointing a heat gun at a DRAM device...) Actual measurements from Google's server farm: * Heat is correlated with increased error rate but does not seem to cause it. Rather, increased utilization drives both. * Hard errors probably dominate error count and drive most uncorrectable errors. * Less than 20% (maybe as little as 8%) of DRAM components accounted for 95-98% of errors. * Initial burn-in works (little infant mortality) but devices start to degrade after 10-16 months. * Little difference in error rate across manufacturers. Increasing size affects error rate in unpredictable ways, often super-linearly. Dr. Schroeder's other work on reliability is also interesting. (2 Comments |Comment on this)