Hacker News Reader

QMLon Who Are My Investors?

Don’t interpret the words given as a common prescription. Most of us don’t have enough to make a global impact—only a local change. Wait and see if Masayoshi Son, or SoftBank, will comment on the situation; see if Stewart Butterfield, or Slack, will. If neither of them say anything, then that’s when you should decide to continue using Slack or drop it; but don’t pretend you were oblivious.

QMLon Ask HN: What's your favorite elegant/beautiful algorithm?

This is also used in TCP with timeouts and ACKs. The idea is you have an expected timeout (ETO) and a retransmission time out (RTO). Initially, you set RTO to ETO and then you send a packet. If you don't get an ACK back, and the timer for the RTO expires, you then set RTO = 2 * RTO. Otherwise, reset RTO = ETO. I think they referred to this as 'exponential averaging' however.

QMLon Ask HN: What's your favorite elegant/beautiful algorithm?

When you are referring to this "variant", are you referring to maximal matching, or bipartite matching when the number of people and schools don't equal?

QMLon Ask HN: What's your favorite elegant/beautiful algorithm?

Doesn't sound too bad; a basic property of XOR is that two of them cancel out, so y ⊕ x ⊕ x is just y.

QMLon Why Mathematicians Can’t Find the Hay in a Haystack

In relation to the article posted, there is a conjecture which states all algebraic, irrational numbers are normal.

QMLon Why Mathematicians Can’t Find the Hay in a Haystack

A more interesting puzzle is to prove or disprove whether every algebraic, irrational number is normal (in some base).

QMLon Why Half a Degree of Global Warming Is a Big Deal

But Celsius and Kelvin have the same unit of measure; they’re only offsetted by 273 degrees.

QMLon Supply Chain Security Is the Whole Enchilada, but Who’s Willing to Pay for It?

Is security considered a “core business function”? I wouldn’t believe so.

QMLon What Made Gödel’s Incompleteness Theorem Hard to Prove

I feel small things like that, or professors not understanding their students’ perspectives — apparently this is a known cognitive bias called “curse of knowledge” — unnecessarily restricts access to field.

QMLon Facebook Network Breach Impacts Up to 50M Users

This is a very typical Silicon Valley mindset where the number of high paying software engineering positions is abundant, and there’s no shortage of work.

If you look outside the Bay Area, I’m sure you will find people who desperate enough and would sacrifice any sense of morals they have for a $100k job.

It’s as simple as that.

QMLon Braess’s paradox

Another relevant discussion, albeit from a more game theoretic view: https://news.ycombinator.com/item?id=17741641.

Beyond two submissions sharing the same subject-title, do you think it'd be possible to find all "relevant" discussions -- say, back-links or forward-links -- as well and post them?

QMLon Show HN: Podflix – Like Netflix for Podcasts

Podcast lists don't load in Safari and in Chrome -- issue for anyone else?

QMLon What happened when Boston Public Schools tried for equity with an algorithm

Before trying Simulated Annealing, try out all the industry grade constraint optimization solvers! They’re surprisingly more effective on average case inputs. I had to learn this the hard way when, for a class project, I was thoroughly beat out by people who just used dedicated solvers for TSP.

QMLon Sloan Kettering’s Deal with Startup Ignites a New Uproar

“The company has an exclusive deal to use the cancer center’s vast archive of 25 million patient tissue slides, along with decades of work by its world-renowned pathologists.”

In general, the world is better off with more information: why not open source the data if better cancer diagnosis is the goal?

QMLon Machine Learning Confronts the Elephant in the Room

Adversarially generated inputs have been known for some time [1], so I wouldn’t call it a new class of attacks unless 1. there are other and older ways of attacking neural networks 2. new as in the last couple of years.

[1] https://blog.openai.com/adversarial-example-research/

QMLon Machine Learning Confronts the Elephant in the Room

That’s actually the concept of ensemble learning! [1] The question is, how would that lead to predicting a randomly placed elephant as an elephant with high confidence?

[1] https://en.m.wikipedia.org/wiki/Ensemble_averaging_(machine_...

QMLon Machine Learning Confronts the Elephant in the Room

That would not seem computationally efficient if for every class (i.e. object) you were to create a 3D modeling and paste it into a training set

QMLon Let Teenagers Sleep In

I wonder if it'd just be easier for high schools to convert over to a college model: students are free to pick courses corresponding to their preferred times, and to leave and enter campus on will. Of course, there will be incentives to consider -- why would a high schooler attend their classes if they're not paying directly out of pocket? In contrast, a school district would try to maximize the attendance rate as that's how funding works (at least in California).

In hindsight, there's always a tension between high schoolers and those who administrate them; it's never a partnership since teenagers are quick to abuse / misuse a privilege.

QMLon Universal Prediction

How would you make an algorithm to predict something unknown?

This is essentially the field of sequential decision making, which encompasses everything from reinforcement learning to multi-armed bandits. [1]

As an example, let’s pretend we have the task of predicting labels of a size 2 alphabet {0, 1}. These labels are come to us via a stream; so in contrast to classical ML, we are not given access to the full “dataset” beforehand to train our predictor. But instead, each round, a label is given to us; and in between each round, we can only predict the next label with the set of labels previously revealed.

Note that we have no model of how the labels are generated — this is what we mean by “predict something unknown”. In general, we can assume two cases: it’s either adversarial or predictable (shaky definitions, I know).

In an adversarial case, we assume the labels are generated such that we incur a heavy loss. Say, if we predict 0, the adversary outputs 1; vice versa. It turns out with deterministic strategies and a certain adversarial model, we can always incur loss linear to the amount of rounds; this goes a bit into game theory. To counter this, we can employ a random strategy where we flip a fair coin to decide what to predict. This is somewhat analogous to you, in the classic tie breaking game, of choosing {rock, paper, scissors} with 1/3 probability.

But what if the labels weren’t generated adversarially? Remember, we don’t have a model of how the labels were generated. Say, the sequence was 000000... predicting with a fair coin would be a bad strategy then; and instead it’d be better to simply predict 0. A better strategy here would be something like follow-the-leader, where you predict the most frequent label seen so far.

But recall, we don’t know how the labels are generated — it could very well alternate between periods of predictability and unpredictability.

What we need is a mixed strategy which does well in both cases. One algorithm that turns out to do this is called multiplication weights, which assigns “weights” to each label and increases or decreases them proportionally each round depending on what the label is.

There’s a whole lot of other ground that can be covered but I’d recommend reading a book instead of reading it here. One resource is “Prediction, Learning, and Games”—but feel free to find another book since this one assumes a background in probability.

[1] Actually the first section of the linked pdf above.

QMLon Driverless Hype Collides with Merciless Reality

From an economic standpoint, investing in fast mass-transit (e.g. high speed rail) would probably result in a larger net positive than self driving cars. Why?

1. I would theorize that the economic benefits of transportation are directly proportional to how fast the service is, and how much people it can move. As an estimation, I'd say cars are limited to 60 mph while high speed rail would be limited to 200 mph.

2. With fixed-path mass transit, the problem of traffic is greatly simplified. Self driving cars may solve a labor issue but they won't solve the traffic problem: cars routing themselves selfishly will not result in the lowest travel time possible. You can argue that a central coordinator can exist to coordinate routes, but if there's an incentive for a person (not necessarily a driver) to deviate from their set path, why wouldn't they take it?

QMLon Driverless Hype Collides with Merciless Reality

we can have a bus system that runs dynamic routes based on the users that are going to ride it and runs closer to ideally efficient vs. a fixed route that is likely inefficient

This doesn't seem to require driverless busses. I can see an argument of why self driving taxis would be helpful -- cost of labor is near zero -- but with busses, the cost of labor is amortized across passengers.

QMLon Burning Man's Mathematical Underbelly

We weren’t always so lucky. One year, before we’d even finished setting up the booth, a cadre of MIT undergraduate physicists stumped us with ‘Soap bubbles physically solve minimal surface problems. Are there any other physical phenomena that quickly solve NP-type problems?’

Anyone have more resources on this? I could only find a brief overview by Scott Aaronson [1].

[1]. https://www.scottaaronson.com/papers/npcomplete.pdf

QMLon Forecasting at Uber: An Introduction

Does anyone know if uber or another on demand company has a similar type blog post, but about predictive matching? Algorithms like stable marriage are fascinating from a theoretical standpoint but I would love to hear about them from people who implement or deploy them!

QMLon How AI changed organ donation in the US

If you go to slide 70 of the CMU link posted above [1], it covers an algorithm called FutureMatch [2], which attempts to assess or predict the quality of a matching. I wouldn't call it AI as the article author calls the algorithm, but more of a subset of ML -- reinforcement learning (RL).

[1] http://www.cs.cmu.edu/~arielpro/15896s16/slides/896s16-14.pd...

[2] https://www.cs.cmu.edu/~sandholm/futurematch.aaai15%20with%2...

QMLon Actuarial science ranked as most valuable college major

I'm fine with metrics, as long as when people say "best", they'll quote an interpretable formula -- you have to start somewhere. The issue with this metric, however, is that there seems to be no given weight towards the population size. What is the point of a metric if it isn't generalizable?

I have the same grip with college rankings. When you look at the top 20, and then at the total student population, you'd realize these rankings are useless for the majority of Americans since only a small fraction will attend these schools.

QMLon My life with 12 programmers, 2 rooms and one 21st-century dream (2016)

Yeah you can find singles for 1k, just make sure you get housemates to share the cost of living space and kitchen.

QMLon The Unlikely Return of Birkenstock

I have two pairs of Birkenstocks, a cheap foam pair and another cork leather one that I just got. I've been wearing them for the past year, even in the winter or in the rain.

My conversion followed from the thought of having sock-less shoes, mainly from the cathartic feeling of removing them after a long day. A cousin recommended me to get a pair of Allbirds; initially, I liked the material of the shoe but it still didn't feel "breathable" (especially on a hot day in summer) and the wool began to rub my feet in a wrong way. This is in addition to the issue of build quality: the glue at the head of the shoe was visible, and also began to deteriorate; I've also had a friend whose Allbirds ripped during a session of basketball -- it lasted only a couple of months.

I'm not sure about how other people feel about this, but if a shoe or a sandal costs over $100, it should last at least for a year or two; I sincerely doubt that the difference between retail price and manufacturing cost is small enough to not cover a replacement 3 times over.

My issues with Allbirds stirred me to find a long-lasting, breathable pair of footwear. I don't remember if it was seeing people on campus wear sandals, or hearing about the longevity about Birkenstocks on reddit that drove me to buying a cheap, "test" pair.

After test-driving a foam version, I decided to invest the mid-tier Arizona sandal (which I'm still wearing today).

Back to the main question, is the "return" of Birkenstocks unlikely. I don't think so: the culture of the company seems to fit well with the demographics of millennials, who aren't afraid to splurge on items with great story / quality behind it.

And anecdotally, the first time I wore them outside, my friends jestingly compared my sense of fashion to a "white, sorority girl" (which I take to be an implication of their popular amongst college students).

QMLon Bull Market Hits a Milestone: 3,453 Days – Most Americans Aren’t at the Party

If compound interest is a force to be reckoned with, shouldn’t we expect power law distributions for wealth?

10% owning 84% seems to be the expectation then. This reminds me of an Atlantic article lamenting the “9.9%” [1].

[1] https://www.theatlantic.com/magazine/archive/2018/06/the-bir...

QMLon A visualization of the prime factors of the first million integers

Are you saying the kth principal component is equal to the kth prime number?

QMLon Slack confirms it has raised $427M at a post-money valuation of over $7B

I've been having trouble coming up with an analogy until now: my main gripe with Slack is that it brings over the concept of open office from meatspace onto digital space. I don't think I'm the only one with this sentiment; it's harder to concentrate on work when notifications go off; even if its not pertinent to me, I feel an urge to stay in the loop on things.