HNHacker News
TopNewBestAskShowJobs

obastani

802 karma · joined September 16, 2013

submissionscomments
obastani··on Why I’m Not a Fan of R-Squared
It's not just philosophical -- there are examples of practical situations where there is a true model (especially in more traditional applications of statistics):

- Polling for presidential election: The "true model" is the voting preferences of all 300 million Americans. A (uniformly) random sample of N Americans can be used to estimate the true model roughly with standard error 1/sqrt(N).

- Particle physics: The "true model" is the decay probabilities of various particles as computed by quantum mechanics. Different models (and parameters) yield different decay probabilities, and experiments can be used to choose between models and/or estimate model parameters.

Of course, oftentimes the true model is philosophical as you describe.

obastani··on Why isn’t the fundamental theorem of arithmetic obvious? (2011)
> So I guess if you're comfortable with modular arithmetic, you can fairly consider this an obvious proof. It relies on another result about primes, but it's very common that one result makes another result easy, and a blanket disallowal of that approach leaves you saying that proving anything is as tricky and non-obvious as proving, proving, that 2+2=4.

But Gowers' point is that FTA relies on a deep fact that is non trivial (as pointed out and proven elsewhere on this thread, the correctness of Euclidean division: https://en.wikipedia.org/wiki/Euclidean_division#Proof). You can sidestep this by using other lemmas in arithmetic, but at some point everything is resting on this one deep fact.

Also, it's not like proving that 2+2=4, which follows trivially from Peano's axioms and the definitions of 2, 4, and +. If something is nontrivial to prove, it's usually because there is something deeper going on beneath. In this case, the fact that Euclidean division works for the integers, but not for other number systems (e.g., Z[sqrt(-5)]), is what makes it deep and interesting.

obastani··on Living Without Atomic Clocks
This is based on my limited physics education, but I believe there are two things at play:

1) Velocity doesn't affect time, acceleration does. So if your twin flies to Mars and back, you age more quickly. But if your twin flies to Mars, and you join him after a year (following the same trajectory), you age the same. This is special relativity.

2) Acceleration due to gravity doesn't count. In fact, the opposite is true -- by standing on the ground, you are accelerating up at 32 feet per second squared. This is general relativity.

So, to answer your question. On the one hand, yes, if I am accelerating (not counting gravity), then I observe time pass more slowly. On the other hand, most of the universe is dominated by gravitational forces, so most systems wouldn't notice these effects.

The fascinating thing is that both (1) and (2) happen because of a very basic physics principle: if I can't use an experiment to tell two frames of reference apart, then the physics in the two frames of reference are identical.

For (1), I observe the same speed of light as someone moving at velocity v relative to me. Einstein used this simple axiom to derive special relativity. For (2), I can't tell the difference between free fall and being at rest (alternatively, I can't tell the difference between being on Earth and being in an elevator in space). This is because the inertial mass equals gravitational mass, a "coincidence" that dates back to Newton's law of gravitation. From this (and a lot of math) Einstein derived general relativity. Beautiful!

obastani··on The Clinton System
Couldn't this be a result of the massively increased amount of conflict in the region since Bush's second term?
obastani··on Why London Underground stopped people walking up the escalators
Actually, it is a problem strictly for the massless :)
obastani··on A new proof of Euclid's Theorem (2006)
It seems to me that the original proof also uses "consecutive numbers are coprime" as the sole property. In particular, it looks at P = p_1 * ... * p_n and says that it is coprime to Q = P + 1; hence, there must exist another prime number.
obastani··on A new proof of Euclid's Theorem (2006)
I think the point is you can easily remove the "deficiencies" of Euclid's proof. Basically, say: let p_1, ..., p_n be the first n primes. We can construct a new prime by examining p_1...p_n + 1. This process can be continued indefinitely.

In fact, I think the statement "this process can be continued indefinitely" is quite misleading -- to formalize this, I think you need proof by contradiction. So, the difference is the syntax used, not the actual semantics of the proof.

obastani··on A new proof of Euclid's Theorem (2006)
The proof isn't referring to n, n+1, and n+2 -- it refers to n, n+1, and n(n+1) + 1. In your example, this would be 2, 3, and 2*3 + 1 = 7, which in fact include 3 prime factors.
obastani··on Leslie Lamport: Computer Science and State Machines (2008) [pdf]
Isn't a stack just a "sequence of terminals and/or non-terminals"? (Or as you suggested, an infinite state machine where the states are all possible sequences of terminals and/or non-terminals.)
obastani··on The C standard formalized in Coq
I think that statement is referring to simply defining "undefined behavior", not actually determining whether a given program has undefined behavior.
obastani··on Deep Neural Decision Forests [pdf]
I think there are a few additional important details.

It's not so much that they stack a random forest on top of a neural network, as that they are trying to train neural networks that have tree-like properties. Deep neural networks can be expensive to evaluate because every layer is needed to compute the resulting label. On the other hand, in a decision tree you only need to evaluate one of a nodes children (i.e., evaluation is linear in the depth of the tree).

They combine these ideas by allowing some nodes in the network to branch (so for a given image, you choose one of two neural networks to evaluate). These nodes don't have to occur at the top level, but can occur in the middle of the network as well.

The difficulty is how you train these nodes; that's where the differential technique for optimizing trees come in. Because you have a unified training algorithm, you can mix and match tree-like nodes and neural network like nodes and train the whole thing.

All in all, the hope is that getting a label for a given node is much faster, so you can run these on smaller devices such as phones without sacrificing accuracy.

obastani··on Quasi-Polynomial Algorithm for Graph Isomorphism
Not really, it's not hard to write a compiler to convert between different programming languages. The challenge is translating between the different semantics of different programming languages (e.g., object-oriented vs. functional, dynamically typed vs. statically typed, etc.).
obastani··on Generating Captions: Describing Videos with Neural Networks
I don't think many serious deep learning researchers claim that it will solve any kind of problems. Instead, it allows us to make significant strides in areas that have been stagnant for many years now (e.g., image recognition, speech recognition, some natural language tasks, and other domain-specific tasks). The potential impact based on these "narrow" domains is already huge (voice commands, image search, analyze medical images, analyze legal/medical documents, aid self-driving cars, etc.). Basically, it allows us to turn data that is highly structured but in a complex way (e.g., images, voice) into standard classification/regression problems. So far, I don't think it does much more than that.
obastani··on The Economist's US college rankings
I'm not saying that each college you attend has no effect on earning power. It's just that I can perfectly predict the effect of each college on your earning power. Does that make sense? If I have an oracle that tells you

"if you got to Harvard, you will make $80,000, if you go to MIT, you will make $86,000",

and the oracle is exactly correct, then under this model, The Economist assigns every college a score of 0.

obastani··on The Economist's US college rankings
That's my point -- who decides what expectations are? Their results are incredibly dependent on the model specification. I imagine if they changed which indicators they used, the results would vary widely.

Here's another way to see my concern. Suppose you had a classifier that achieves 1.0 R^2; then since it perfectly predicts each school's expected value, it'll assign each school a score of 0. I'm pretty suspicious of an approach where the results get worse with better predictive power.

Even if you want to do "exceeds expectations", I think you shouldn't include variables that are school specific, only variables that are student specific. In other words, for my expected outcomes, which school is best?

obastani··on The Economist's US college rankings
This analysis doesn't make any sense to me. Even by their own intent (i.e., best added value based on salary), their methodology is nonsensical.

For example: Why should CalTech get hurt for being near L.A.? They're basically arguing that you're better off going to a school in the middle of nowhere because "hey, for being in such a crappy location, you did pretty well!". In an absolute sense you are better off going to CalTech, it's just that they might not leverage their advantage as well as some other schools.

Not that they even show the last point -- it seems unlikely that the true model is linear (I'm guessing they used linear regression). For example, if the true model is closer to a sigmoid, then schools at the high end suddenly get unfairly penalized and schools near the low end get unfairly boosted.

Finally, the statistical indicators are equally misleading. I can obtain an R^2 of 1.0 just by including indicators I[is COLLEGE_NAME] for each college. While that might not give you significance, the point is that getting good prediction is meaningless.

I think what they really want is to restrict to predictors about the students. So, given that you're a straight A student with a 2400 SAT score, what would you expect to make coming out of each school? This at least tells me something about the added value to me of going to a certain school. (This approach is still prone to bias, but in the opposite direction -- there's a chance that the straight A student with a 2400 SAT score going to community college may have been smart but unmotivated, which might correlate with lower salary.)

Edit: Here's another concern. They're claiming to have a model for "expected" earnings:

earnings = A * (college covariates) + b + error

but they can't distinguish between model error (i.e., error because their model is misspecified) vs. the school variation that they are trying to capture.

obastani··on Tesla autopilot stops a collision [video]
I don't know about about your city, but I would expect drivers to stop for pedestrians even if they are crossing in the middle of the road. The reason I never cross in front of a moving car is a matter of courtesy and not wanting to break the law, not out of fear of being run over.
obastani··on Obama Encryption Policy Rejects Laws Mandating Backdoors
What if there was a way to give only the US government access to the data? For example, somehow have a shared encryption scheme where the user holds a decryption key to decrypt her data, but the government and the company can somehow use their keys in conjunction to decrypt the data as well. (This means company must be complicit in releasing the data, so they have the opportunity to verify/fight the warrant.) If someone came up with such a secure encryption scheme, would that be a palatable approach?

I ask this because I'm curious whether you (and people in general) believe that hackability is the only obstacle to giving government access to personal data (with a warrant, of course), or if people would still be uncomfortable with such a system.

Personally, I think such a scheme would be a good compromise, but I'm not a crypto expert so I don't know if it's in any way feasible.

obastani··on The economics of software correctness
Yes, I was referring to the verified parts. To the best of my knowledge, these are bug free -- I believe the bugs were found in an unverified portion of the front-end (which I think was subsequently verified). I think the technique works in principle; it's just very, very expensive.
obastani··on The economics of software correctness
I don't think that's a fair analogy. At this point, I believe we have pretty good ideas that make it possible to write correct software. Most impressively, there has been progress on creating a completely verified C compiler named CompCert: http://compcert.inria.fr/. This really means "completely verified"---they formalized the semantics both of C and of assembly, and used an automated theorem to prove that the semantics of the generated assembly are the same as the semantics of the source code.

Of course, building (and verifying) CompCert took something like 10 years, so it's certainly not a feasible way to write software yet. Maybe a better analogy would be around the invention of the first airplane---we have a long way to go before everyone's flying in a jumbo jet :)

obastani··on Big Tech Has Become Too Powerful
One potential issue is data -- Google can leverage its huge number of historical searches to improve search for everyone. I doubt this is a huge issue for players such as Microsoft, but it probably poses difficulties for a small start-up.
obastani··on Self-driving cars can be disabled with a laser pointer and a Raspberry Pi
The design is very much intentional. A huge barrier to adoption for self driving cars is perception of safety; Google wanted a car that looks safe not dangerous.
obastani··on It’s probably a myth that we’re sleeping less than we used to
The claim made by the article isn't that getting < 8 hours of sleep is bad (they agree it is), it's that most people get 8 hours of sleep; more specifically, that many surveys overestimate the number of people who get too little sleep.
obastani··on Registered clinical trials make positive findings vanish
It's fine if you run a separate experiment to justify the hypothesis. If you choose the hypothesis after conducting the experiment, then the p-values you obtain for that hypothesis are invalid.
obastani··on Speed matters: Why working quickly is more important than it seems
I think there's a big difference between working quickly and working rushing a project. Doing things thoroughly and correctly is of course well worth it; you can complete this work as efficiently as possible without cutting corners and building half-baked products.
obastani··on A Beginner’s Guide to Restricted Boltzmann Machines
My understanding is that deep learning has the largest benefit with perceptual data (e.g., images, speech, and to a lesser extent natural language and control). The driving force behind deep learning is the ability to "compress" data by learning hierarchical representations. For example, your images of cats and dogs are very high dimensional and encode a lot of redundant information, making them especially suitable for such compression.

On the other hand, deep nets probably won't outperform more traditional machine learning algorithms on less high dimensional, uncompressed data. This problem is exacerbated by the amount of tuning needed by deep nets. Architectures have been fine tuned for common data types such as images and speech, but if the dataset doesn't fall into one of these categories, you have to tune the algorithms yourself.

obastani··on Startups and The Big Lie
Does that apply after the options have been exercised? Can you have a contract saying that you'll share part of the payout from exercising the options?
obastani··on Why we're all-in on virtual reality
I've tried thinking about how virtual meetings would work, but to me it seems to be flawed. Virtual reality allows us to view a 3D environment as if we are there, but how would we capture that 3D environment to begin with? Is there technology that allows us to capture not just 3D, but 3D from all angles? I expect that for "presence" to work, when you move your head, the perspective that the camera sees (on the other end) would need to change, and I can't imagine how this would be achieved.
obastani··on Users Rationally Rejecting Security Advice (2009)
The article is talking about total economic costs (i.e., to all users, the bank, etc.), not just from the perspective of the banks. In other words, we as a society may lose more by implementing "best security practices".
obastani··on Tetris blocks traumatic flashbacks even after the memory is fixed
Their control group seems a bit questionable to me -- maybe sitting there quietly for 12 minutes increases the risk of flashbacks? It seems possible that you'd end up thinking about the traumatic event.
← PreviousPage 6 of 7Next →