|
| |
|
| |
|
|
|
|
TCHS 4O 2000 [4o's nonsense] alvinny [2] - csq - edchong jenming - joseph - law meepok - mingqi - pea pengkian [2] - qwergopot - woof xinghao - zhengyu HCJC 01S60 [understated sixzero] andy - edwin - jack jiaqi - peter - rex serena SAF 21SA khenghui - jiaming - jinrui [2] ritchie - vicknesh - zhenhao Others Lwei [2] - shaowei - website links - Alien Loves Predator BloggerSG Cute Overload! Cyanide and Happiness Daily Bunny Hamleto Hattrick Magic: The Gathering The Onion The Order of the Stick Perry Bible Fellowship PvP Online Soccernet Sluggy Freelance The Students' Sketchpad Talk Rock Talking Cock.com Tom the Dancing Bug Wikipedia Wulffmorgenthaler       |
|
bert's blog v1.21 Powered by glolg Programmed with Perl 5.6.1 on Apache/1.3.27 (Red Hat Linux) best viewed at 1024 x 768 resolution on Internet Explorer 6.0+ or Mozilla Firefox 1.5+ entry views: 2198 today's page views: 241 (8 mobile) all-time page views: 3386282 most viewed entry: 18739 views most commented entry: 14 comments number of entries: 1226 page created Fri Jun 20, 2025 05:47:38 |
|
- tagcloud - academics [70] art [8] changelog [49] current events [36] cute stuff [12] gaming [11] music [8] outings [16] philosophy [10] poetry [4] programming [15] rants [5] reviews [8] sport [37] travel [19] work [3] miscellaneous [75] |
|
- category tags - academics art changelog current events cute stuff gaming miscellaneous music outings philosophy poetry programming rants reviews sport travel work tags in total: 386 |
| ||
|
It was rewarding to finally be able to sleep in after setting up experiments, in what hopefully will be my final year as a graduate student, and then feed the hamsters slowly, one in each hand, the pleasure therein only heightened by the fact that they can do nothing for me. What more can a guy want?  Not much at all (Source: Somewhere on the Internet) Another of my lab seniors got recruited by Google, with a quite impressive junior interning with Facebook next semester (note: the Programmer Competency Matrix is a fair gauge of coding ability); which reminds me that I'm not far from having to think about a Real Job™, despite all my best efforts. Almost neglected to mention my primary school math teacher, who got mentioned in TODAY at the beginning of the month - both the name and the little red bridge fit. Come to think of it, a number of my past teachers have had good arms, with a History tutor-cum-rugby coach sporting some great guns, though surprisingly it was the more sciencey types who were better at making books fly. Exciting days. It was also nice to note that the Public Service Commission has apparently gotten the missive that the longtime educational rat-race will likely remain unchanged despite the PSLE tweaks, as they finally begin to select for diversity. A drop in the ocean is better than nothing, I guess. Then again, the relevant Minister remains adamant that tuition is not necessary for good grades, in the grand incumbent tradition of being technically correct but realistically missing the point, while the more tabloidy The New Paper hit it on the head. Meanwhile, the outlook for our Top Investment Firm is bright, after they more or less got awarded another local monopoly. No word for loss in their dictionary! It's not so rosy for Apple (for now), with their continued refusal to put their name on anything cheaper than about US$450 pricing them out of emerging markets. Brand preservation might not even be a bad strategy, other than the problem of their new covers inducing trypophobia in some. Well, they've got a ton of reserve cash to burn through. Rounding it up, former Presidential candidate Tan Jee Say may be hopping to the Singapore's People Party, and while it may not have the heft of the WP, NSP or SDP, will probably allow him a shortcut to the front. If so, it's encouraging for potential participants to set out their stalls early, instead of engaging in the last-minute hopping that is a turn-off to so many voters, from recent elections. Do Or Do Ta  For more first blood! Luxury splurge of the month was this SteelSeries QcK Mini DotA mousepad, replacing the Razer Kabuto that was getting icky. One immediate plus is that it's not shifting as the Kabuto was wont to, though it's less cooling to the touch. Here, I must take offence at those who dismiss DotA as "just a game" - from picking heroes, to what to do (and not to do), there's a lot of cognitive load involved... at least if you wanna win. But that was the whole point, right? Boys just want to have fun, and kick ass, prefably at the same time. Additionally, (almost) nobody knows who anybody else is, which refreshingly allows all to duke it out on pure merit. Oh, and it's even educational, if you know where to listen, and what to delve into. It's getting big! Butt-whupping got a bit tougher in Diablo 3 after Blizzard announced through email that they will be shutting down their item auction houses, perhaps after realising that allowing cash-flush players to buy their way to dominance might not make for a satisfying experience for the bulk of their base. Haven't touched Oboham for a year, anyhow. Mixed Signals Slightly pumped by last week's read, which seems to have filtered into the wider consciousness pretty quickly going by how this TODAY commentator asserted that Singapore must be antifragile, I popped by the library for some related works, and a familiar name caught my eye - Nate Silver, it turns out, has written The Signal and the Noise: Why So Many Predictions Fail - But Some Don't. For those who haven't heard of Silver, he came to prominence during the last two U.S. national elections, for running a blog that correctly predicted the winner of 98% of the state-level elections beforehand. This got him on resident Master Political Analyst Herr Ahm's radar, and although he would never admit it, I suppose that he learnt a thing or two from his approach. Make no mistake, it's hefty - over 450 pages, plus another eighty-odd for the notes and index - and for good reason. It covers most of the well-worn basics and more besides, so for newcomers to the area, this would make a decent standalone introduction to practical probability. [N.B. There appears quite a bit of cross-fertilization going on, with The Black Swan acknowledged in a footnote on page 368. Super Crunchers, which was next to it on the shelf, got a blurb on the cover by a Freakonomics author] Again, skimming through SparkNotes style: we can't predict some things well or at all (sub-thesis of Antifragile), many experts can't either, we have so much data now that more and more of it is not relevant (noise), averages are bad (very Talebian), but we can improve our predictions in some fields (like hard science... and chess, where the computer is already king), so let's see how it works in poker, financial markets and climatology; and remember, there are unknown unknowns. While we're at it, the borders of physics seem to have been pushed back, with researchers finding a method to express Feynman diagram calculations in geometric terms; as my knowledge of this field is approximately nil, the best analogy I can draw is that of computation in a transformed space, and also that simplification is key. Of course, not all are convinced that it's that big a deal, with one wit going as far as to lampoon it by noting that when a structure is unrestricted enough, it surely describes the desired phenomena... and everything besides. Then again, he's a computer scientist. We're kinda good at this sort of thing. We might as well follow on this by mentioning that the National Security Agency, having found time off from building Star Trek replicas, has come to the startling realisation that terrorists love Gmail. If they had continued to dig deeper, there might be a small chance that they would also have made the groundbreaking discoveries that Wal-mart cashiers love Gmail, suburban soccer moms love Gmail, college students love Gmail, military personnel love Gmail, Republicans love Gmail, and even that, wait for it, humans love Gmail! Tragically, it transpires that @ikeelyou.net is getting no love. Fitting this into the book's theoretical basis (which only comes in at Chaper Eight, probably under sound advice that visible math causes sales to plunge, as communicated to Hawking; indeed, analysis of frequently-cited law review articles in Super Crunchers notes that those that included an equation were cited a lot less), the evidence that a person uses Gmail has next to no impact on the posterior probability of whether he is a terrorist. But when since has logic stopped security agencies from wanting to scan everybody's correspondence? (Silver does later show that even 9/11 was not an outlier) At The Heart - Statistical Engineering Anyhow, a good bit of that chapter is largely devoted to the struggle between Bayesian and frequentist (Fisherian) probability. The main distinction appears to be that frequentists (attempt) to be completely objective, in the sense that any errors are purely due to not sampling the entire population (i.e. underlying parameters are assumed to be fixed and constant). Bayesians, however, submit that the actual mean - the ideal form - does not actually exist, and must remain unknown. Thus, since only the data is real, the mean can be expressed as a probability. Further, Bayesians feel justified in using an initial estimate, or prior probability, that is possibly divorced from the data, if required.  (Source: xkcd.com - Explained) Silver holds that pure frequentist statistics is limited due to its disavowal of context, bringing up Fisher's real-life rejection of the seemingly incontrovertible relationship between smoking and lung cancer following his methods as an example. In particular, he states that admitting a prior estimate is valuable, if only because it allows us to recognize our biases, and monitor how new evidence can possibly affect our beliefs. Another strike against the frequentist paradigm is how untenable it often is in actual use - knowing that Manchester United's long-run winning percentage is 60%, what can I draw from this about tonight's clash with City? Clearly, the parameters are not constant - even if I were to use only the head-to-head record, quite a few players would have left the teams since the last time they met, and further the abilities of the remaining players would have changed in the meantime. It should be noted that Silver got his start as a baseball stat geek, a sport that he notes has been amenable to statistics because of the relatively low dependencies between personnel - a good pitcher or batter will perform mostly independently of his colleagues in his individual stats, unlike say in football, where having Messi won't do too much good against an average professional side, if his teammates are merely of Sunday league standard. He further describes a very successful basketball bettor, whose world-class prediction rate is... 57%. As it turns out, this roughly 4% profit margin after accounting for the bookmakers' cut was enough to let Voulgaris make millions - but it took him probably thousands of hours of watching NBA games, with his dedication extending to following players' Twitter feeds, all for that tinest of edges. Finishing just above the money with far less prep doesn't seem too bad now. Voulgaris is also a poker star, and on this Silver describes the formation of an online poker bubble (which he personally indulged in) around 2003. The underlying economics, to him, is obvious - the existence of suckers. Since no new money is created at the table, we can infer from first principles that for there to be winners, there must be losers. The distribution, from data mining, is what Silver supplies, and his observation is that the worst players (bottom 10 percent) lose an average of $400 per hundred hands at $5/$10 tables, while the best ones (top 10 percent) win only slightly more than $100 per hundred hands. It might be wondered how this can be, since in the same way no money is destroyed at the table, and the answer is that there is a relatively heavy churn of lousy players - they lose a thousand bucks or two, and then quit, while the very best guys soldier on. Note that even with these suckers, most of the above-average players were still barely getting by. Therefore, when legislation to regulate online gaming got passed and drove out many bad players, what happened was that the difficulty ratcheted up, to the point that even the top 30th percentile of players were losing money over the long term. Former sharks became suckers. However, Silver adds that even the best of the best can crash, since poker is intrinsically very volatile; charting the probability intervals for a very good player who is expected to make $200 per hundred hands in the long run, Silver reveals that after 10000 hands, he would still have a perhaps one-third chance of being in the red. Indeed, Chris Moneymaker, who did so much to popularise poker with his $2.5 million fairytale win in the 2003 World Series, is earning just $110000 a year before accounting for tournament entry fees, which while not to be sniffed at, is probably comparable to what a fairly run-of-the-mill accountant would be approaching. There's a slightly more sinister side to all this, with Super Crunchers reporting that some casinos are modelling the optimal "pain point" for each customer, as defined as the maximum they can lose while still being willing to return to lose even more. When they detect that a player has reached that point, a "luck ambassador" will sidle up and offer some freebie as a token of appreciation. Aren't they nice? At The Hart it may begin to lose its value as an economic indicator. - Goodhart's Law Silver asserts that economic data is very noisy, far more than that in say meterology. This goes back to the beginning of the book, where he criticized collateralized debt obligations (CDOs) as being flawed, due to an inherent assumption that the risks of default by mortgages within each CDO are independent. It is easily shown that, by adding some correlation (which probably exists), the probabilities of default quickly show little resemblance to the original estimates. Fama's No Free Lunch (i.e. efficient market hypothesis, you can't beat the market) and the Winner's Curse (I did say the book is a comprehensive survey) are also in, as well a a more novel take on part of why there are bubbles (because we herd), and following that, why we herd. The outcomes of a financial trader are listed:
Basically, there's a strong incentive to go with the flow and collect money when the going is good. Even if things go south, well, it could be the other guy who gets eaten. Whereas by being a downer bear, the rewards aren't that great even if one turns out to be right. Best to keep the music going. What I thought was unsaid was that bubbles facilitate (unearned) gains, and furthermore are self-sustaining in that at some point, it could become rational for people to participate in the bubble, even if they do not believe in the underlying fundamentals. On the first point, we can imagine a bubble-less economy. Here, all permanent assets appreciate at about the same rate. Then, there would be no incentive to speculate, since no investment could give more than the average (and standard) rate of return. This is, of course, also completely unrealistic, since external factors will inevitably cause some assets to become more (or less) valuable over time. The problem happens when these adjustments gain a momentum of their own, as observed countless times, perhaps most prominently in real estate; Fama, true to his hypothesis, states that "if you can't tell you're in a bubble, it's not a bubble", but history shows that few who were in a bubble, acknowledged it as it built up. Part of the explanation is due to it being more socially acceptable to trust in the good times, along with shorting the market - the anti-bubble bet that in theory checks exuberance - being highly risky. We further observe that power and knowledge can only be exploited in an economy with persistent mispricing (i.e. bubbles). As we have seen, there is no way to pull ahead through purely financial means in a no-bubble world. However, with bubbles, a method to multiply one's wealth reveals itself - get in early, get out near the peak, rinse and repeat with the next bubble. The only problem that remains is identifying bubbles, but with sufficient (and not unrealistic) muscle, it is not too difficult to... proactively arrange matters, and completely legally at that, at least a good proportion of the time. Once any player (or cabal of players) gets fat enough, all bets are off. In the worst case, even the sensible are basically forced to commit - imagine that a bubble has formed, somehow, for decorated toilet paper, and that in the past year their prices have risen ten-fold, with no signs of stopping. If the volume of transactions is large enough, even those who recognize the value of toilet paper for what it is may figure that they could be proven right eventually, but by this time Joe Schmoe next door could have parlayed his $10000 in savings into a million dollars and cashed out, while they would still only remain on $50000, and we can't have that, can we? The only way to get even would be to join the party, and since they're so smart - at least, smarter than Joe - they'll surely exit in time... right? The figures may not be as dramatic in most instances, but the mechanism remains. Much of wisdom consists in cutting through bullshit Of course, this can't go on for ever as long as the money supply remains fixed, but seeing as how the Fed is continuing to churn out dollars (still waiting for a major media outlet to de-euphemize "quantitative easing" by just saying "money printing"), it could drag on for awhile more. There's the obligatory story about Lorenz and chaos theory, where minute initial differences result in very different outcomes in dynamic systems. More interesting is the revelation that the issue is being resolved by perturbation and repetition... wait, isn't that just noise and model averaging in machine learning? And isn't group forecasting/the wisdom of crowds, that was claimed to reduce forecast error by 15-20% in many disciplines, simply committee networks, or more generally ensembles/boosting by combining weak learners? [N.B. Super Crunchers has an overview of neural networks] (Opening, Up) This fashion for "don't know" has gotten into academia as well, going by recent work on clamping the half-spaces of support vector machines (SVMs) by T.E. Boult (introducer of the Bachelor of Innovation degree at UCCS), who dropped by NUS last week. Essentially, in the standard binary SVM, decision boundaries are created, but at the end the output is simply one of the classes (possibly, but not commonly, with attached probability). However, and with good reason, Boult reminds that we often do not have access to all the possible classes in actuality, and therefore need a way to just throw our hands up and say "I dunno", when the situation warrants it. Apparently, this approach met with a wee bit of resistance from the establishment, but the intuition - that we should not be confident of predictions where data is sparse (again the extrapolation headache), which Silver would undoubtedly agree with - seems solid. Assuredly, it is a rather large step to convert this basic idea into actual usable formulae, which the paper does, with overgeneralisation and overspecialization risks balanced against training data (prior probabilities), extending pure single-class density estimation/novelty detection techniques. They go on to demonstrate it with cute squirrels, which Mr. Robo puts down as a definite plus. Next: Side Write
anonymous said... Q."What more can a guy want?" A.co a sensual cock rub through the jeans followed by a mind numbing blowjob
gilbert said... Sir, that was intended as the device known as a "rhetorical question".
anonymous said... the only device i know is vibrator
gilbert said... Sir, I suggest extending your range to literary devices - the pen is mightier, after all.
anonymous said... if the pen is very sharp and the sword is very short...LIKE YOUR PENIS HHAHAHAAHAHAHAHAHAHHHAHAHAAHAHAHAHAHAHHHAHAHAAHAHAHAHAHAHHHAHAHAAHAHAHAHAHAHHHAHAHAAHAHAHAHAHAHHHAHAHAAHAHAHAHAHAHHHAHAHAAHAHAHAHAHAH FUCK YOU
gilbert said... Sir, in all honesty it is shorter yet, for I have not yet gone beyond the word; but if you should throw an "s" upon it, our exchange could then well be complete.
anonymous said... aiyah just fuck your hamster lah fuck until shiok shiok
gilbert said... Sir, they are my current squeezes.
anonymous said... so usually u one cock u put how many hamster on it
gilbert said... Sir, I cannot help but be impressed by your ability to squeeze that way.
|
|
||||||||||||||||||||||||||||||||||||
 Copyright © 2006-2025 GLYS. All Rights Reserved. |
|||||||||||||||||||||||||||||||||||||
