[3-Week Course] "Deep Learning from Scratch 1" Completion Challenge with the Translator

If you want to start deep learning "properly" but have always hesitated in the face of formulas and terminology, let's try a different approach this time.

This reading challenge is a 3-week journey to read <Deep Learning from Scratch 1 (Remastered Edition)>, where we will understand the principles by "implementing core concepts from scratch" rather than relying on frameworks.

This book starts with the perceptron and moves step-by-step through neural networks, learning (loss functions and gradients), backpropagation, and training techniques, ultimately leading to the basics of Convolutional Neural Networks (CNN) and image recognition. Complex concepts are explained through visualization with computational graphs, and the practice is designed to turn "knowing" into "doing" by following along with hands-on Python coding.

✅ Here is how the Completion Challenge works

Every week, you will read a set amount of material (theory), execute and modify key code yourself (practice), and reflect on what you've learned by writing a brief review. The goal is to take a study process that often fizzles out when done alone and push through to the end "together" by setting specific weekly goals. The program is structured so that you can follow along even if you aren't very familiar with Python.

✅ What are the participation benefits?

• Hanbit Mileage Benefits: Up to 20,000 Hanbit Mileage points will be awarded to those who complete the course.

• Translator Q&A: There is an opportunity to directly ask the translator questions about points where you get stuck while reading, getting a feel for terms/concepts, or questions like "Why is this implemented this way?"

• Opportunity to attend an online seminar with the translator: "You've built your foundation—where should you go next?"

✅ Recommended for these people

• Those who want to start studying deep learning with <Deep Learning from Scratch 1 (Remastered Edition)>

• Those who want to understand deep learning principles through implementation rather than just "memorizing" them

• Those who want to properly organize core deep learning concepts such as CNN and backpropagation once and for all

✅ Tips for a successful challenge

• "Working code" comes first: Even if you don't understand the formulas, type and run the source code provided in the book to see if it works first. Backtracking while running the code will help your understanding.

• Don't give up on backpropagation: Week 2's 'Backpropagation' section will likely be the most difficult part. I recommend trying to understand it by drawing the computational graphs yourself.

• Utilize the Translator Q&A: If you get stuck while studying alone, make active use of the Translator Q&A. Specific questions like "Which part of the code corresponds to this formula?" are recommended.