2.1.1 Python Fundamentals

Why Python?

Every IOAI task comes as a Python notebook, and every library you will use, from pandas to scikit-learn to PyTorch, is a Python library. So before any machine learning, you need to be comfortable reading and writing basic Python.

The good news: you do not need to become a software engineer. IOAI code is usually short. You load some data, loop over it, call a few library functions and print the results. The contest even gives you a small LLM for help with code (see the LLM usage policy). But the LLM cannot think for you. You still have to read its code, spot what is wrong, and change it. That is exactly what this module prepares you for.

If you have not installed Python yet, do Setting Up Python first. Or skip installing for now: the main course below runs in your browser.

What you need to know

By the end of this module, you should be comfortable with these ideas:

That list is shorter than it looks, and every course below covers all of it.

Pick one main course

There are hundreds of Python courses online. Do not collect them. Pick one, finish it, and do its exercises. Here are the ones we recommend, and who each suits.

Our pick: Kaggle Learn Python

Kaggle Learn Python has seven short lessons. Each lesson is a notebook to read, followed by an exercise notebook where you write the code yourself and check it instantly. Everything runs in your browser; you only need a free Kaggle account. It is also written by data scientists, so the examples already feel like the work you will do at IOAI.

Do the lessons in order. The table shows where to read more if a lesson feels too fast.

For more depth: CS50P

Harvard's CS50P is a full university course, free online. Each week has a lecture video, written notes and a problem set. It goes slower and deeper than Kaggle, and it is a great choice if you have never programmed before or if you like learning from lectures.

For this module, weeks 0 to 4 are enough: Functions and Variables, Conditionals, Loops, Exceptions, and Libraries. Week 6 (File I/O) is useful later, when you start reading data files. Week 8 (Object-Oriented Programming) belongs to the next module.

If you prefer something else

• A book: Automate the Boring Stuff with Python is free to read online. Chapters 1 to 8 cover this module, and Chapter 5 is a friendly introduction to debugging.
• Lots of small exercises: the University of Helsinki's Python Programming MOOC checks your answers automatically. Parts 1 to 4 are plenty for now.
• One long video: freeCodeCamp's Learn Python: Full Course for Beginners gives you a quick overview in one sitting. Follow it with one of the courses above, because watching is not the same as writing code.

A quick tour

Here is what the ideas look like in code. You do not need to understand every line yet; come back after your course and check that you can. Try typing the examples into a notebook and changing them.

Variables and types

name = "Asha"
age = 16
accuracy = 0.87
passed = True

print(type(age), type(accuracy), type(name), type(passed))
# <class 'int'> <class 'float'> <class 'str'> <class 'bool'>
print(f"{name} is {age} and scored {accuracy:.0%}")
# Asha is 16 and scored 87%

The f before the quotes makes an f-string: anything inside {} is replaced by its value. :.0% shows a number as a percentage with no decimals.

Lists and loops

scores = [72, 85, 90, 64]

total = 0
for score in scores:
    total = total + score

print(len(scores), total, total / len(scores))
# 4 311 77.75
print(scores[0], scores[-1], scores[1:3])
# 72 64 [85, 90]

Positions start at 0, so scores[0] is the first item. scores[-1] is the last one, and scores[1:3] is a slice: items 1 and 2, but not 3.

A list comprehension builds a new list in one line. This one keeps the scores of 70 or more:

high_scores = [s for s in scores if s >= 70]
print(high_scores)
# [72, 85, 90]

Conditions and functions

def grade(score):
    if score >= 85:
        return "A"
    elif score >= 70:
        return "B"
    else:
        return "C"

for s in scores:
    print(s, grade(s))
# 72 B
# 85 A
# 90 A
# 64 C

Python uses indentation (the spaces at the start of a line) to know which lines belong to the def, the if or the for. Keep it consistent: four spaces is the usual choice.

Dictionaries

A dictionary stores key: value pairs. Here it counts how often each label appears, which you will do all the time with real datasets:

labels = ["cat", "dog", "cat", "bird", "cat", "dog"]

counts = {}
for label in labels:
    counts[label] = counts.get(label, 0) + 1

print(counts)
# {'cat': 3, 'dog': 2, 'bird': 1}
print(counts["cat"])
# 3

counts.get(label, 0) means "the value for this label, or 0 if it is not there yet".

Libraries

import math
import statistics

print(math.sqrt(16))
# 4.0
print(statistics.mean(scores))
# 77.75
print(statistics.median(scores))
# 78.5

math and statistics come with Python. Later you will import bigger libraries in exactly the same way, such as import pandas as pd in Pandas.

Reading error messages

You will see errors every day. Everyone does, including the people who wrote this guide. The skill is not avoiding them; it is reading them calmly. When Python stops, it prints a traceback. Here is the one you get for asking a four-item list for scores[4]:

---------------------------------------------------------------------------
IndexError                                Traceback (most recent call last)
Cell In[1], line 2
      1 scores = [72, 85, 90, 64]
----> 2 print(scores[4])

IndexError: list index out of range

Read it from the bottom up. The last line says what went wrong: position 4 does not exist, because a list of four items only has positions 0 to 3. The lines above it say where: Cell In[1], line 2, with an arrow pointing at the line that failed. This is how an error looks in a notebook; in a .py file the layout differs a little, but the last line still tells you what happened.

These are the errors you will meet most often at the start:

For more, see Errors and Exceptions in the official tutorial and CS50P week 3.

How to practise

• Type the code yourself. Copy-pasting skips the part where your fingers and brain learn the syntax.
• Change the examples. What happens with an empty list? A negative score? Breaking code on purpose teaches you a lot.
• A little every day beats a lot once a week. Thirty minutes a day for two weeks is better than one eight-hour Sunday.
• Ask for help the right way. Try help(len) in Python, search the official documentation, or ask in the SOTA Discord. If you ask an LLM, ask it to explain rather than to write the code for you; the prompt in How to study works well.

Check yourself

Try each one before opening the answer.

If these felt easy, you are ready to move on. If not, no problem: go back to the matching Kaggle lesson and redo its exercise.

What comes next

• Object-Oriented Programming, Classes, and Inheritance: the class keyword you will see in PyTorch code.
• Simple Python Problems: more practice with small problems.
• Jupyter Notebook Interface and Basics: the notebook format every IOAI task uses.