patrick-llgc/Learning-Deep-Learning

Overview

This project is a personal collection of reading notes on deep-learning and machine-learning research papers. It is maintained by Patrick Langechuan Liu, who works in AI for autonomous driving at Nvidia. The repository contains:

Paper notes — Summaries and commentary on many research papers, organized by month and year. Each entry links to the original paper on arXiv or a publisher page, and some include a separate note file (e.g., paper_notes/vlash.md) with the author's own analysis.
Reading suggestions — A list of recommended first papers for newcomers to computer-vision deep learning, plus a list of trusted sources for finding more papers.
Blog posts — Links to the author's Medium column “The Thinking Car”, covering topics such as planning, BEV perception, drivable space, and object detection in autonomous driving.
Podcast notes — A single interview note from a podcast episode with a researcher at OpenAI.
Scratchpad — Quick, informal notes (like a gist) on specific technical topics, currently covering compute hardware and attention masks.

Author

Patrick Langechuan Liu — a physicist by training who transitioned to deep learning and autonomous driving. He currently leads ML modeling for Nvidia’s end-to-end autonomous driving project, Alpamayo.

Contents Details

Paper Notes — The bulk of the repository. The notes are grouped under headers like 2026-04 (1), 2025-12 (0), etc., showing the upload date and number of papers in that batch. Each entry gives the paper title, a link, and often a short tag (e.g., [Async infer], [latent CoT]) or a conference badge (e.g., <kbd>ICLR 2026</kbd>).
Review Posts — Links to ~20 Medium articles, each addressing a specific subtopic in autonomous driving or deep learning. Several posts also link to related paper notes in the repository.
Suggested Reading — A dedicated file (start/first_cnn_papers.md) plus the author's own notes on that reading list (start/first_cnn_papers_notes.md).
Extras — A short list of “trustworthy sources of papers” (trusty.md) and a podcast-interview note.

Structure

start/ — beginner reading material.
trusty.md — source recommendations.
topics/ — notes that accompany specific blog posts.
paper_notes/ — individual paper notes (many are linked from the main list).
gist/ — scratchpad notes.
podcast/ — podcast note.

The repository also has a minimal webpage generated via GitHub Pages (linked at the top).

Purpose

This is a personal knowledge base for the author, but it is publicly available and likely useful for others interested in deep learning, particularly in computer vision and autonomous driving. It provides a curated entry point for beginners and a detailed archive of recent research.