Snow in Pasadena

11 minute read Views

“It’s your golden hour.” — JVKE

City of Gold

People like to take photos during golden hour, which usually right before the sunset.

The last few days in Pasadena, every hour of sunlight has been golden hour.

Sometimes, it even snows. Ashes, but still pretty.

First Snow

Starting on Tuesday 7th, massive gusts of wind woke me up and there were torrents of leaves flying past my window.

Later, when I went to class, I thought it was a bit chilly and I talked to J about it. He said that we should definitely ask M, a friend who is quite the expert in meteorology.

M said it was just the Santa Ana winds.

Second Wind

Apparently, there were 60 mph winds coming down the mountain from the north. Naturally, this knocked over a few power lines and caused some fires.

At some time after dinner ~18:30, it smelled like grill night but it was unclear where the smell was from, but I could see some smoke in the north.

This was apparently the start of the Eaton Fire which spread east and west in Altadena around 8 miles and north and south 5 miles, covering 10.6k acres by the 8th, then 13.7k and 14k the following days.

I asked my brother whether the fires would reach the school and he replied that it was impossible because there just aren’t enough trees for the fire to jump to.

However, if the wind picks up again like it’s expected to, things may actually get really bad.

Crowd Control

At Caltech, most people have left to the south in Irvine or Orange County. Some people have even escaped to the Bay Area or the east coast.

The main reasons to escape include the uncertainty of water contamination and air pollution as a result of burning buildings with asbestos, benzene, lead, and all sorts of delicious chemicals.

In the midst of the chaos, there have been lootings and even more arson. While the cause of the big fires are unknown, it may be a mix of arson and the wind.

Incredibly, E ordered a lot of masks the day before the fire simply because he doesn’t like to wear sunscreen. Thanks to that, we had masks immediately when we needed it.

Fourth Wall

Typically hiding behind a screen whenever there are bad things happening in the news, I’ve actually gone to the other side, a victim of nature’s plot.

If you look at an aerial photo of the Pacific Palisades, it appears to be as bad as various warzones. As a longtime player of Enlisted, the scenes reminded me of France from WW2 with buildings stripped to the bone.

Arcanist

Some student took an amazing photo from the top floor of Caltech Hall and it looks exactly like Heimerdinger’s memory of the world collapsing from people using magic.

alttext View of Altadena from Caltech Hall

How to survive the apocalypse

  1. Get some bottled water
  2. Wear a mask
  3. Don’t go outside

This is sounding an awfully lot like 2020 all over again, huh?

CS42 Notes on Teaching

I’m taking a course CS42 on how to teach math and science and I have to do a weekly discussion question and take notes, so I guess I’ll just write it here:

How children learn:

  • Babies prefer clear images to blurry ones, kind of clever experiment where they rigged pacifier suction level to the focus of a TV
  • Babies prefer learning new things due to habituation, when they get used to something they lose interest.
  • Children have bad memories, so use chunking

Computational thinking:

  • Everyone should learning how to think like a computer scientist
  • Q: How can everyone learning to think like a computer scientist if there are not enough people who understand computational thinking in K-12 education to teach it? For instance, 170k students took AP CSP in 2024 but there are 17 million high school students in America, so 1% of all students are learning computational thinking.

Learning objectives:

  • Objective before content
  • Students should be able to ___
  • Remember, understand, apply, analyze, evaluate, create

Cooking ingredients

Meat: Pork Vegetable: Maybe asparagus or bean sprout stirfry

What really happened

I bought some flank steak from trader joes along with some asparagus, bell pepper, and an onion.

For lunch, I made some rice (with way too much water) and stirfry beef + bell pepper + half an onion.

It wasn’t that great since I didn’t remember soy sauce for the stirfry and I also had to eat C’s leftover rice from a restaurant because my rice was so disgustingly mushy.

For dinner though, I made some adaptations with the use of soy sauce and putting less water in for the rice. I clearly didn’t put enough water because by the time the bottom was getting dry the rice was still hard.

What do boys do?

I noticed that a lot of my classmates had a very boring childhood so I would like to explain some of the things that my friends and I did for fun.

Sounds

Whoever could make the strangest sound was highly respected, for example, I remember this kid Z was really good at making dolphin sounds, and another kid C who could burp the alphabet. Really impressive stuff.

There were also a variety of fart simulations like the armpit technique, along with the lesser-known kneepit variation.

Actions

There’s a funny thing you can do with a hoodie which is putting one arm at your stomach and the other holding the empty sleeve, then moving them up and down in alternating fashion to create one of the most striking, almost inhuman, movements.

I did this to J yesterday and he said “I haven’t seen that since elementary school!”

Tag is one of the great games of all time with the freeze and zombie variations.

More notes on Teaching

The annoying thing about teaching kids is that they can’t pay attention for more than a few seconds to something you say. This is really bad if you are trying to teach them a complex idea.

What kinds of practice and feedback enhance learning

Practice should be focused on the goal, like playing a whole piano piece vs focusing on just the hard part. Goal-directed practice + targeted feedback. Feedback must be linked to another opportunity to apply the skill.

Explanatory Power of the Hierarchy of Student Needs

Students need to feel safe emotionally and intellectually. They also need belonging/esteem which only few teachers can achieve. Some kind of inside joke seems like a good bonus, and I think DMO does a lot of things that help with student needs.

Inclusivity and equity are also mentioned as parallel to the goals of satisfying student needs. However, I wonder whether it would be correct to personally request a high-performance student to ask dumb questions to encourage insecure students to ask questions.

On an individual basis, this student is no longer equals with their peers, but other students have been led to believe that this student is also capable of asking bad questions.

Why Do Student Development and Course Climate Matter for Student Learning?

Student’s level of development/maturity interacts with classroom climate.

The Chickering model suggests many dimensions like competence, emotions, autonomy, identity, purpose, integrity.

“Climate works in both blatant and subtle ways, and many well-intentioned or seemingly inconsequential decisions can have unintended negative effects with regard to climate.”

This tip would suggest that teachers cannot ensure every student’s needs are met, but they can improve the learning experience by adapting to students’ level of development.

In one of the examples, the professor does not call on women students in lecture because he doesn’t want to put them on the spot, which results in patronization. However, in a K-12 setting, it’s important to not put too much pressure on insecure students by singling them out because they may become even less confident in the future.

Q: If students are supposed to be treated equally, how can teachers shrink the gap between the best and worst students without treating them differently?

Responses

I posted this question on the course forum and a student responded that “best” and “worst” are bad labeling and I agree because I don’t really agree with this framing in the first place, that’s why I put asterisks around them.

However, this student immediately said afterward that an example of how to resolve the problem is the way Caltech gives some first years supplemental math and writing courses

AI

I was quite surprised to learn that data augmentation/hallucination is equivalent to regularization. I guess it makes sense because you are making the

Resnick Sustainability Institute

I recently attended a talk by a JPL Chief Scientist of some science directorate and it was super boring and I will probably not go to one of those talks again.

Essentially, the guy was talking about how he was asking people at JPL to write 3 pages each for his 90 page report on how climate change is going to influence the lab itself and LA county.

It seems completely pointless to write this kind of report because it’s just going to become a stack of paper on some officials’ desk which will inevitably get thrown in the trash after it’s never opened.

Even worse, this guy was more excited about the prospect of “holding a printed copy” of the report than helping anyone affected by the fires. By the way, he has a $400K budget for this report and 1000 copies are going to be printed and not publicized.

There’s a lot of

Cognitive Dissonance

where he wants to appear to make an impact (he mentions things like how the 2028 Olympics are going to be hosted in LA) but it’s also a report that people will never see or care about because it’s not going to be available to the public.

When one of the audience members (I think it was some hotshot at RSI because he’s at all of these talks) asked a question relating to wildfires, the first thing he responded with was how “successful” the webinar from a few weeks ago was, since it had 2000 people.

Then, he mentioned that after the fires happened, he was in talks with a coauthor who said that it may be better to change the whole report to just be about wildfires, but he quickly dismissed that idea because of all the time that already went into writing the paper (two months as of then).

So, basically he’s basically applying the sunk-cost fallacy because he would rather keep going on a doomed path where thousands of dollars have already been wasted than find better alternatives. Or maybe it’s actually impossible to have 30 people write 3 pages each in 2 months.

Anemoculus Collector

Genshin is a game that I have disliked for many years, but I finally tried it and it’s actually quite fun. Essentially, you do quests, explore the world, fight monsters, do puzzles, and lootbox pull new characters.

I got lucky and pulled two five-star characters within 20 pulls, so that got me playing a bit too much every day.

Satisfying Parts

The fighting is quite satisfying initially because there are many elemental combinations that do bonus damage, so you can use one character of water type and attack with fire type later to vaporize the enemy. This teaches kids phase transitions pretty well, I guess.

Some parts of the story are incredibly interesting and the environment is very engaging, especially in the region based on Atlantis. The art quality is super good and it really immerses the player.

Unsatisfying Parts

If you pull a newer character, then you need to go far from the original continent to level them up which messes up the feeling of the intended progression, and sometimes it feels like you’re going places just so you can teleport to them later.

The stamina system is incredibly annoying because you can only run for a few seconds before you have to recharge and walk. If you run out of stamina while climbing a mountain or swimming, you fall/drown.

Some of the puzzles are super confusing because there is no explicit goal, it’s kind of like those physical puzzles where you are trying to open the box but there’s no rules given. I guess your enjoyment depends on how much you like that but for me it’s not that satisfying.

Mining ore is also annoying because you need a claymore (big sword) character to mine anything but you may have a better character so there’s no point in having that character in your party.

Genetic Algos for Ensemble Blending: to DMO

I’m taking a machine learning course where we form teams and compete against each other to make the best model for some task like song popularity classification and my team had a pretty wild performance from last place 50th to 3rd to 26th.

My friend accidentally selected a terrible model from day 1 and that was averaged with the really good model made with genetic algorithms. We would have been 14th if there was no accident.

There’s a concept of blending model predictions together with a weighted average which can decrease variance in the predictions and hopefully predict better on the testing data.

So, I wrote a GA to crossbreed different sets of weights for 5 different models and the results were better than all the other stuff we tried.

RBF: Applied ML

The RBF kernel is the radial basis function kernel which can be applied to support vector machines for classification machine learning problems.

It also stands for something else which some of my friends used to describe my typical disposition.

Re: Genetic Algos for Ensemble Blending

DMO wrote back a few days later that they couldn’t find a new CS teacher and that there’s only one semester of Applied Algorithms now, which is very unfortunate, but there are always ups and downs I guess.

Leave a comment