My way of learning atcoder (~green)

Written on 6/25:. After doing atcoder for about 5 weeks, I was able to verbalize what kind of learning I am doing, so notes Let’s read it back and tweet it when it turns green. →Tweeted with tidy additions as it turned green on 7/5.

• Enter the contest first
  • Because learning after testing is better for memory retention. see Testing is a means of memory.
• Write with the intention of explaining to others how you solved it.
  • Because it strengthens my memory and gives me the opportunity to organize and understand better. see teach others.
  • While I was writing the explanation, I thought, “Hey, I thought it was A, but that’s not necessarily true, is it?” or “I messed around and finally decided to write AC this way, but looking back, I wonder if this part of the code is not necessary. In retrospect, this part of the code may not be necessary.
• View the official explanation PDF
  • There could have been a better approach to the problems I was able to solve.
    • I can look at it after I solve it once and say, “I see!” I see!
    • Compared to reading a textbook, it has a stronger sense of “self” and is more easily absorbed into the brain.
  • Know the policy for problems you couldn’t solve.
    • If you can now say, “I see!” then solve the problem again.
• See the code of strong people
  • Especially when you say, “Hey, the algorithm shouldn’t be wrong, but it TLEs.
    • Search for “AC code in Python for that problem,” sort by speed, and look from fastest to slowest.
    • Now I’ve learned how to use Numba.
  • There was one short code search.
    • E of ABC171.
    • I thought, “There must be a more concise way to write.”
    • But I haven’t done it since, because the code hit me with a golf-like readability dump, and it’s hard to read.
  • I’m not aware of the author at first, but as I repeat myself, I notice “someone I’ve seen a few times”.
    • When information is recognized not by itself, but by the “person who created it” behind it, the source of information can expand from there
    • For example, Python’s fast code ranking often has maspy
    • I happened to see a blog by maspy, and was glad to find a detailed explanation. It’s called “Finding quality sources of information”.
    • This experience creates a learning behavior of “let’s see if maspy’s blog has written about this”.
  • When the code written by others is not clear at a glance
    • Save it locally and use vscode’s variable renaming function to rename it to an easier-to-understand name as you read it.
• Refer to others’ code for implementation.
  • Don’t look and copy, don’t look and write.
  • 1: Read it, understand it, and say “I see, that’s how you do it”, then write it yourself without looking at it.
  • 2: If you can’t write it, look at it right away.
  • I’m spinning fast in a little cycle called
  • This is the effect of doing “small tests at the level of a few lines” first, which has the effect of “testing first reinforces memory.” see Fast test cycles.
  • For example, when you are writing, you may think, “Huh? Does this subscript need to be -1? Do I need -1 for this subscript or not? I think it’s necessary in my opinion” and write it, and then I look at your example and confirm, “Oh, sure, it’s -1. Someone else’s code says, “Good, good, good! You’re headed in the right direction!” The image of a voice saying to you, “You’re on the right track! It’s easy to stay motivated because the rewards are given in a short period of time.
  • At this time, prepare the variable names and code for your future reading self.
    • Use variable names that are easy for you to understand, not those written by others
    • Change the writing style to one that is easy to understand
      • For example, if I see something written with ternary operators and I think it’s hard to understand, I write it with if statements to make it easier for myself to understand.
      • I’d supplement the omitted else clause and add, “In this condition, you don’t have to do anything because
    • When your future self thinks, “Oh, isn’t this the same problem I solved before?” and reread it, the code should be easy to read.
      • I don’t know if there will ever come a time to read it back, but this rewriting process in itself helps me understand it better.

7/6 I’ve noticed something else since writing the above article.

• Verbalization of Awareness
• Different people see the same problem statement and start implementing different programs.
  • ABC172C is a problem set up to take one from each of two mountains A and B.
  • The most naive form is to implement what is written in the problem statement exactly as it is written, but this often TLEs
  • I saw this problem, and I was unconsciously transformed.
    • When I saw someone solving it naively, I realized, “Okay, I was deforming unconsciously.
  • I was curious what caused this unconscious transformation.
    • So I started experimenting with naming the techniques used in the deformation process and making them tags on Scrapbox.
    • In the case of this problem, “route-independent”
      • A then B, B then A is the same.
      • Then there is no need to distinguish which came first.
        • (As I write this, I noticed that it is related to “Unordered columns are multisets,” and like this, after the fact, “Oh, this and this are related,” and a network of knowledge is created in the brain.)
• This is something we’ve been trying for a while now, so I can’t quite explain what the effect is yet, but it feels beneficial.

7/11

• I was solving DP problems Educational DP Contest without any hesitation, but they were getting harder and harder, and when I saw the sign that said “If you AC all of them, it is equivalent to yellow,” I felt I made a mistake in selecting the difficulty level.
• I found that I can see the recommended difficulty level in atcoder problems, so I look at that.
  • easy is indeed easy, moderate is about “easy if you notice it, but it feels difficult because you don’t notice it”, solving this seems like a good idea. #atcoder
• How engineers learn
• way of learning

---

This page is auto-translated from /nishio/僕のatcoderの学び方(〜緑) using DeepL. If you looks something interesting but the auto-translated English is not good enough to understand it, feel free to let me know at @nishio_en. I’m very happy to spread my thought to non-Japanese readers.