Trains of Thought and Rambling. Low-effort Mini Essays.

This section is just a place to put my unfinished ideas and rambles, and shouldn’t be taken very seriously. Nor should you expect anything here to be comprehensible or coherent, and definitely not useful.

If you find a typo anywhere in here, please make a PR to add a fix to it! You can also feel free to add your own opinions on any of these subjects and append them on a PR, with a link to your blog or source if you want.

Overdoing it as fault tolerance

This came to me due to my current Japanese studying schedule of 1.30hs a day, every day of the week. My smart goal is to keep learning 15 new words from JLPT3 deck every day.

However due to a thing called life, every once in a while a thing would pop up and I’d be busy or distracted, and would not be able to dedicate the allocated time for Japanese studies that day. This mostly happened on days where I saw people or had Discord parties (lockdown times make us creative).

My solution? Move the Anki bar so it’s 18 new words a day. This way, since 618 > 715, I can skip a day a week if I feel like I won’t make it (or will have to squeeze it too tight on my schedule, preventing relaxed study), and still make my original quota.

This is nothing new, companies do this with KPIs all the time: overshoot so you hit the mark. But I’d never thought of what this implies for fault tolerance: if every day I’m overdoing it a bit, then when a random event brings down a piece of my infrastructure or resources, I can still survive.

It’s like stocking up for the winter!

Do bees like celery? On randomly lost knowledge.

So today I was walking to get food and thinking: the api in apiculture -should mean bee- sounds a hell lot like apio (celery). And apio totally sounds Latin, like it should’ve been Apium or something way back when. So I wonder if bees are the apio-eating insect or whatever, and that was very common knowledge in ancient Rome but now a random person like me wouldn’t know it.

A quick check in the wiktionary entry for abeja says:

From apis (“bee”) because of bees’ perceived inclination towards the plant.

Seems like my guess was correct! But now I have to wonder. What other things are embedded in ancient names, and how many are true knowledge like this one, and how many just lies? Man, I love etymology.

Continuing the etymology walk, I was surprised today to learn that wrens belong to the family Troglodytidae. That is to say children of troglodytes. Why? Wikipedia says “The family name Troglodytidae is derived from troglodyte, which means ‘cave-dweller’. Wrens get their scientific name from the tendency of some species to forage in dark crevices.” Interestingly the Spanish name is fully formal as troglodítido, though vernacular names abound (matraquita being by far my favorite).

So 文句, 文字, 文明 all start with the same kanji, but it’s read different every time. Namely, もんく、もじ、ぶんめい respectively. Seriously, nipponjin? That’s today’s rant.

The word “bear” actually means “the brown one”. The actual word for a bear is lost to history; nobody would say it because they truly believed that saying the real name would summon an actual bear. So all we have left is the non-magical euphemism.

The same happens in Russian where they use medved which means “Those who know honey -med-“.

Bears were the first Lord Voldemort.

When checking out a new hot unsupervised learning algorithm, something the Berkeley course guys do is test it with a 1d pair of variables. Like “oh does this GAN/distribution alignment thingamagig work? Let’s imagine X and Y are 1d random variables and see how that would work”. I take this as a very good first principles approach to understanding the algorithm, and would extend it to 2d for visualization too. I just think reducing example problems to small ones is a nice trick I keep stumbling upon in many different areas and it would be good to have this mapping of “for X domain, typical small problem is Y” like how in graph theory you always start induction with a K1 graph, or in DCS you use a small LTS with maybe 3 states.

[Share on twitter]

21 Nov 2020 - importance: 4