Preparation
Even though we had planned to participate a month in advance, we didn’t start preparing until the week before the topic was released, and in some cases, even just one day before. So I think we were essentially going into the competition “blind” or unprepared. I want to outline our degree of preparation in three parts:
- Model: few commonly used models
- Write: basic frame
- Program: some code frame of commonly used models (most learning models), haven’t tested
Process of competition
I was the writer in our team.
Day 1
Morning: Started reviewing the questions two hours after the problem statement was released and spent approximately one hour on it. Since historically the C question in previous competitions could often be addressed using deep learning models, and our programmer had a good understanding of deep learning, we focused our attention on reading the details of the C question (and choose it finally).
Afternoon: Discussed some critical terms. Tried to understand tennis rules. Programmer started coding. Writer did nothing.
Night: Programmer “addressed” the problem 1 and 2 with 70% accuracy . (Sadly it was “wrong”, and we did not find that until the day 3)
I felt I was playing the role of modeling. We thought MCM was just so-so. (As it turned out, we underestimated it) Went back to sleep early.
Day 2
Morning: Writer and Modeler tried to find some related article.
Afternoon: Wrote some not important part of our essay.
Night: Used some tricks (it was wrong too) that improve the accuracy to 90%.
Felt happy.
Day 3
Morning: We found we seemed used a wrong method.
Afternoon: We chose to solve these problems using a method with 60% accuracy.
Night: Drew some pictures, wrote our solution of problem 2 and first part of problem 3.
Felt anxious and nervous.
Day 4
Morning: Drew some pictures, wrote our solution of problem 3.
Afternoon: Tried to improve our accuracy but failed. Wrote our solution problem 5.
Night: Continued writing solution of problem 5.
Late Night to early morning: Wrote the other problems, integrated them together.
Extensively utilized LLM (e.g. ChatGPT) for assistance in writing and translation. We ultimately submitted our paper two hours before the deadline.
Analysis
Team
- Advantage:
- quick and simple communication due to our team members were gathered together
- programmer have the knowledge of machine learning
- modeler is an even-tempered person
- writer can use LaTeX
- others
- Disadvantage:
- everyone is contributing ideas, but there was no single independent authority directing decisions.
- have no experiences of this kind of competition
- time arrangement
Myself
- Advantage:
- LaTeX
- English reading
- other
- Disadvantage:
- always put huge pressure on team members
- have no experience and related knowledge
- English writing and translation
- other