Where does Taiwanese air pollution come from?

December 16th, 2015

Conclusion: The pollution now (December) is from China, the pollution in November was locally generated.

Earlier this year, there was a bout of posts about whether the air pollution in Taiwan came from China. The air is apparently pretty bad right now, and this time I think it’s clear that it’s from China. I just pulled this screenshot from waqi.info:


Screen Shot 2015-12-16 at 2.02.24 PM

Observe that the pollution in Yilan Country, southeast of Taipei, is not that different from Taipei itself. This is what you would expect if the pollution came from far away. Compare this to the map from November (http://www.thenewslens.com/post/243421/):



Notice the huge disparity in homogeneity. Also, if I remember correctly, back then the wind was also unusually still. I think based on these facts, I would conclude that the pollution in November was locally generated and the pollution now is from a far away source, probably China. (Disregard the low pollution directly to the West, the wind doesn’t come from there. Also, there could be a time lag, so even the pollution levels in the direction of the wind source at the same point in time are not immediately relevant.)

If I had all the historical data in a csv, could probably do a better analysis than two snapshots, if anyone could point me to a source I’d be happy to do that.

What parts of Deep Learning are modern?

December 14th, 2015

Conclusion: outside of a very brief period in which pre-training with Unsupervised Learning was shown to be helpful, Deep Learning has largely been about hardware brute force, and learning how to use brute force to solve problems.

Terms I need to learn more about

  • Pattern Deformations

  • Hessian-free learning

  • Batch Normalisation (Thanks A Breitman)

  • Competing Units

Read the rest of this entry »

Deep Learning with Small Data

December 12th, 2015

I looked into the topic a bit more, and found this exchange, which I think makes sense to me.

The essence of the argument is that because Google etc. have a lot of data, they develop techniques that can make use of that data. However, if you do not have a lot of data, there are other, maybe less developed, techniques to use.

Neal Stephenson on The Problem with Personalised News Feeds

November 3rd, 2015

From The Diamond Age:

That the highest levels of the society received news written with ink on paper said much about the steps New Atlantis had taken to distinguish itself from other phyles.

Now nanotechnology had made nearly anything possible, and so the cultural role in deciding what should be done with it had become far more important than imagining what could be done with it. One of the insights of the Victorian Revival was that it was not necessarily a good thing for everyone to read a completely different newspaper in the morning; so the higher one rose in the society, the more similar one’s Times became to one’s peers’.

News has an important social function, and it’s often more important to read the same news as your peers than it is to read the news that is the most interesting to you.

Movies – Inside Out, The Little Prince

October 31st, 2015

Inside Out was amazing because it didn’t have a villain. In contrast, The Little Prince portrays all “normal” adults as villains. I never noticed this when I read the book years ago. Youthful angst is a frustration with being forced to obey rules that one can’t understand. Prince proposes resolving that dissonance by looking inward, and denigrates the customs of others as senseless and not worth interacting with. Contrast this French approach with the Japanese one in here.

Learning to accept unknowability

October 25th, 2015

Conspiracy theories and superstitions have the same origin. They are both attempts to deny the pervasiveness of randomness in life. People who become overly invested in a low-noise worldview are prone to late-life conversions to superstition because they are so invested in the idea that the world is predictable that they would rather switch hypotheses on the basis of noise (and hence overfit) than admit that the signal-to-noise ratio is that low.

Admitting Unknowability is much more terrifying than admitting Unknownness.

Wittgenstein’s ladder

June 20th, 2015

Austrian Philosopher Wittgenstein once described the structure of his expositions as such:

My propositions serve as elucidations in the following way: anyone who understands me eventually recognizes them as nonsensical, when he has used them—as steps—to climb beyond them. (He must, so to speak, throw away the ladder after he has climbed up it.)
He must transcend these propositions, and then he will see the world aright.

This concept is known as Wittgenstein’s Ladder (wikipedia: do read this)

A lot of the finesse in designing a modern syllabus lies in understanding how to construct this ladder, such that

  1. The first rung is reachable from where the student is right now
  2. Each following rung is reachable from the previous rung
  3. The final rung is where you want it to be, and goes far enough

Knowing where you want the final rung to be may not tell you very much about the first rung at all, because the first rung could be completely fictional relative to the last – what’s important isn’t consistency per se, but the ability to conceive of a continuous path of rungs in between them. The presented facts can outright contradict each other, even, if that helps promote faster ladder-climbing.

Lagged asset correlations, a thought experiment

December 1st, 2014

Suppose I start a fund that imposes a 1-day withdrawal lead time, and takes your money and invests it in the S&P 500 on day 0, but then reports the day 0 return as the day 1 return, the day 1 return as the day 2 return, and so on, reporting the return on day 0 as 0. This fund has a return which is a tiny bit less than the S&P, but is completely uncorrelated on a daily basis. It would, however, have an almost perfect correlation on an annual basis, and that’s why you wouldn’t invest in such a thing. If you only used daily correlations when deciding whether to put this asset into your portfolio, you would have been greatly mislead!

If you optimized for track record Sharpe ratio you would actually invest in this stupid fund. That TYPE of smoothing is worthless though!

The legitimacy of capture

October 29th, 2014

EconTalk on regulatory capture and economists: http://www.econtalk.org/archives/2014/10/luigi_zingales.html

Incentives are important to think about, because getting incentives wrong can mean pitting another human against you, and by symmetry there is no telling who would win when that happens.

Immediate material gains are a part of incentives which are relatively easy to understand. What’s discussed in this podcast under the label of capture is the fact that there are systematic influences that go beyond immediate individual gain. Regulators tend to come to sympathize with the regulated, because in their shared knowledge and social circles they are closer to each other than they are to the rest of us.

This is a very tasty idea when you think about how any ideology can interpreted as capture. To be ideological is to have an opinion based on a framework, more so than on objective data. Often this is the right thing to do, because data is too noisy in individual cases to be leaned on too heavily (http://xkcd.com/1132/).

I think the solution is not to try and be neutral – that is impossible. Rather, we should declare our roles in order to detach our egos from them. We do so by speaking in such manner, for example about Uber:

  • speaking as a libertarian, I think people should have the right to use whatever tools they want as long as it’s a transaction between willing parties
  • speaking as a consumer, I like that I can call cars that are cheaper than cabs
  • believing in the rule of law, I think that Uber is violating the spirit of taxi medallions, and the state should step in to either compensate the cabs for a lost property right, or enforce those rights
  • speaking as an anti-monopolist, I think that Uber is getting too powerful

It would be nice to have a list of these tropes, similar to tvtropes.org


September 15th, 2014

As many of you know, I left Katong Capital as of May this year, where I worked with Yishen Kuik and Jeff Ma for the last 3 years. The statistics was fun, the infrastructure building was fun, and I have had the fortune to continue to do those things at my current job with Minus Inc.

I still greatly enjoy finance ideas, the same way I will continue to enjoy electronic structure. I also have a lot more to say about working in small partnerships now, hah. Ideology, can’t live without it, can’t live with it – it can be difficult to diagnose insanity from the inside!