Mola-maker's Study Vault

A public log of what I'm learning: math, ML, ADAS for two-wheelers, algorithms, and English (CET6). Stu in JLU

About me

Undergraduate student. Strong interest in math (October competition prep) and ML/LLM fundamentals. Currently building an ADAS system for two-wheel vehicles. Fervantly in learning everything about Coding

What's in this vault

• daily/ — daily learning logs
• 00.Machine Learning Note/ —ML course notes, paper notes, fast.ai,Git usage
• (waiting for establishment)
• 10-math/ — calculus, linear algebra, competition problems
• 20-code/ — Codeforces solutions, algorithms in C++/Python
• 30-ml/ — ML course notes, paper notes, fast.ai
• 40-adas/ — ADAS project log: design, bugs, decisions
• 50-cet6/ — CET6 vocab and timed-section practice
• 90-meta/ — weekly reviews, plans, reflections

Current focus (May 2026)

• Math final on July 1
• Daily Codeforces in C++, no AI-generated code
• ADAS perception pipeline
• Python + NumPy fundamentals (CS231n)

How to read this vault

Browse folders directly on GitHub, or clone and open in Obsidian for the full graph view and link navigation.

Contact

Feel free to open an issue or fork. Always happy to talk to like-minded learners.

Suggestion Claude gives

Python — skip the generic tutorials

You don't need another "Python basics" course. You need Python for the things you'll actually do. Two paths:

For ML/scientific Python specifically — the gold standard:

• CS231n Python+NumPy tutorial (Stanford, free): https://cs231n.github.io/python-numpy-tutorial/ — read in one sitting, then type out every example yourself. This is the single best resource for going from "knows Python" to "can do ML in Python."
• NumPy official quickstart: https://numpy.org/doc/stable/user/quickstart.html — short, dense, official.

For general Python depth (so you stop being scared of code):

• "Automate the Boring Stuff with Python" (free online): https://automatetheboringstuff.com/ — this is the one to actually do exercises from, not just read.

• ML papers — the realistic beginner reading order

The Quora answer in my search is actually right that throwing papers at beginners often backfires. Papers assume context. So here's the order that works: Quora

Step 0 (before any papers): one solid course

• Andrew Ng's "Machine Learning Specialization" on Coursera (audit free) — still the best on-ramp.
• Or fast.ai's "Practical Deep Learning" (free, https://course.fast.ai/) if you prefer top-down/code-first. Given your "I can't read code" problem, fast.ai is probably better for you — you'll be reading and modifying real PyTorch from day one.
• Step 1: foundational papers (read in this order)

1. "A Few Useful Things to Know About Machine Learning" — Pedro Domingos. Not a research paper, more a wisdom dump. Domingos provides a comprehensive overview of essential machine learning concepts and common pitfalls — a great starting point for understanding the broader landscape of machine learning. Read this first. ~10 pages. GeeksforGeeks
2. LeNet-5 (LeCun et al., 1998) — birth of CNNs. LeNet-5 was the first architecture capable of learning features automatically, without manual feature engineering. Medium
3. AlexNet (Krizhevsky et al., 2012) — the paper that started the deep learning era.
4. ResNet (He et al., 2015) — introduced residual connections, allowing the training of extremely deep networks by mitigating the vanishing gradient problem. Still everywhere in modern architectures. Jobs-in-data
5. "Attention Is All You Need" (Vaswani et al., 2017) — the Transformer. The single most important paper of the last decade. Don't read this until you've coded a small neural net first.
6. Batch Normalization (Ioffe & Szegedy, 2015) — short, you'll see it in every codebase.
7. Step 2: directly relevant to ADAS (read after step 1)

• YOLO v1 (Redmon et al., 2016) — "You Only Look Once: Unified, Real-Time Object Detection." Since you mentioned YOLO, read the original first, then jump to whichever modern version your project uses.
• U-Net (Ronneberger et al., 2015) — for segmentation, useful for lane/road detection.

Step 3: only after the above

• Karpathy's "Let's build GPT" video series — code a transformer from scratch. Then DeepSeek's papers and code will make sense.

Where to find them: all the above are free on arxiv.org. Just search the title.

One curated GitHub list to bookmark (not read all of):

• https://github.com/hurshd0/must-read-papers-for-ml — the "must read papers for Data Science, or Machine Learning / Deep Learning Engineer" list. Use as a reference, not a checklist. GitHub