Posts

Showing posts from July, 2018

Algocynfas advances!

We can now do minimum spanning trees.

Next up: binary search trees.

David Attenborough, empty posh accent

I was alerted to the fact that this "naturalist" apparently doesn't care what nonsense they give him to read when I was binge-watching the "Wild [X]" (where X = "India" or "China" or "Indochina") series, and heard him describe a encounter between two red pandas as "a male driving off a male intruder," and in a subsequent episode, describe the exact same footage as "a male trying to attract a female." (I actually went and re-played the earlier episode to make sure.)

In Blue Planet, Attenborough treats the viewer to such gems as:

"Spring comes as the sun begins to climb higher in the sky."

Nope, by the time spring comes, the sun has been climbing higher in the sky for three months!

"It is now mid-summer, and the days are getting longer."

No, mid-summer's day is when the days begin getting shorter. And if he means "the middle of the summer," the quote is even worse.

"Kelp, l…

Great paper on computer science education

Here.

Racist dog whistles

These days the progressive left talks continually about "racist dog whistles" being sounded by... well, any figure they don't like.

Here's the thing about dog whistles though: only dogs can hear them! And that is the idea behind the "racist dog whistle" meme: this is secret, coded language that only racists can understand. But...

That means the progressives pointing out these "dog whistles" would have to be... racists! Or else they couldn't hear the whistle, they wouldn't understand the code.

Of course, what the charge really amounts to is, "I don't like that person's preferred policies, and while I can't point to anything explicitly racist that he said, I can instead smear him by claiming he's speaking in a secret racist code."

More sportswriter math silliness

Here, from Zach Lowe:

"There is no shame in that. Cousins can't make the Warriors much better because it is mathematically impossible for a team this good to get much better."

Mathematically impossible? Does Lowe not realize there are an infinite number of positive integers, all of which it is mathematically possible for the Warriors to score? And that all of the Warriors opponents scored more than zero points in every game they played, so they could do much better on defense as well?

I think what Lowe really might have meant was "statistically unlikely," but that doesn't sound as dramatic, does it?

Statidummies

Sports announcers current infatuation with statistics and probabilistic reasoning is only matched by their ignorance of any fundamental statistical principles. Consider this passage, answering how many championships LeBron James will win with the LA Lakers:

"Pelton: If I had to bet on one outcome, I'd probably say zero. The Warriors are still around, and other challengers are forming. I still think going to the Rockets or the 76ers would have given James a better chance of winning a championship. That said, the average outcome for James is probably closer to one championship than none."

So this gut thinks the "average" number of championships James will win is closer to one than zero, but he will "bet" on zero. You can bet he won't bet on this at all, and that he cites "statistics" just to look like a "modern" sportswriter who is all probabilistic and whatnot.

Back to PropArgs Again

This was an early instance of "infrastructure as code."

I am now rebuilding "PropArgs" in a new context, that of agent-based models. Here is the order I think we want properties to have precedence, from lowest to highest:


The database: this is where to store fairly static info on the model being run, as well as server names, etc.Environment variables: allows overriding DB settings for dev / prod servers, or particular runtime environments, such as web, GUI, or command line.A stored property file: can create settings for particular runs, e.g., to do big batch runs of a program with specific settings.The command line: can override previous properties by invoking the program with a novel setting.Asking the user: the ultimate authority on what parameters to use for a model.

Any thoughts? Any place to set these parameters that we've missed?