Start here

This project is a learning and paper-reading guide for machine learning, artificial intelligence, computer vision, NLP, large and multimodal models, and computational pathology. It is designed for readers at different stages, but its deepest route is designed to support research-level work with whole-slide images (WSIs).

The resource combines a curriculum map, a central curated paper-reading guide, detailed background curricula, a dedicated natural-image Vision & Language field, nine specialist computational-pathology chapters, a filterable WSI Data Atlas, a daily arXiv Radar, and a reader-controlled library. Curated routes use human-reviewed metadata; automated discoveries remain clearly separated until they are reviewed.

Looking specifically for papers? Start with the central paper-reading guide, then use How to read a paper to audit each paper consistently. Use the background literature registry to filter the canonical ML, CV, NLP, LLM, and natural-image vision–language records. The Daily Radar is a provisional discovery feed, not the curated reading list.

Want to record your own reading? Open My Library. Opt in to Google Drive to resume the library across browsers and devices. Browser-only storage and portable JSON remain available as independent alternatives, and calendar reminders can be exported separately.

Choose a route

Undergraduate route

Start with ML and deep-learning foundations, then choose computer vision or natural-language processing. Move into natural-image Vision & Language after learning both image representations and transformers. The goal is to learn the vocabulary, implement a small model, and read one paper without treating every equation as a prerequisite.

Master’s route

Review the foundations, study transformers and self-supervised learning, then reproduce a paper with a clean train/validation/test split. For natural-image multimodality, follow image–text representation → grounding → multimodal LLMs → evaluation rather than starting from a demo. Choose a staged sequence from the paper-reading guide and use How to read a paper to turn reading into testable questions rather than a collection of summaries.

Computational-pathology research route

Follow this sequence:

1. Computational-pathology overview
2. WSI engineering and quality control
3. Cell, gland, and tissue modeling
4. Weakly supervised slide learning and MIL
5. Self-supervised representation learning
6. Foundation encoders
7. Pathology vision–language models
8. Spatial, molecular, and outcome modeling
9. Evaluation and clinical translation

The background routes are not detours. They supply the concepts needed to judge whether a pathology method is genuinely new, correctly evaluated, and clinically relevant.

Use a repeated learning loop

For each topic:

1. Orient: read the learning objectives and define unfamiliar terms.
2. Understand: connect the intuition to the formal objective or architecture.
3. Implement: reproduce the smallest useful experiment.
4. Interrogate: identify assumptions, leakage risks, and failure modes.
5. Recall: close the source and explain the method from memory.
6. Apply: decide what would change on a new cohort, organ, stain, or scanner.

A paper counts as understood when you can reconstruct its main claim, method, evidence, and limitations—not merely recognize its abstract.

Trust and scope

Core curriculum items are human-curated. Radar entries are automated discoveries and remain visibly provisional until reviewed. A link is not an endorsement, and no model will be presented as universally best: results depend on task, cohort, preprocessing, adaptation, and evaluation design.

This repository links to primary papers, official documentation, and lawful resource pages. It does not copy third-party figures or mirror paper PDFs. Figures are original or clearly attributed and license-compatible.

The project is educational, not clinical guidance. Do not use its content, code, or model notes for diagnosis or patient care without appropriate validation, governance, and regulatory review.

A good first week

• Day 1: select a route and write down your current prerequisites.
• Days 2–3: study one concept and implement a minimal example.
• Day 4: read one foundational paper using the three-pass method.
• Day 5: explain the paper without notes and record the gaps in My Library.
• Day 6: inspect the data split and evaluation protocol.
• Day 7: write one falsifiable follow-up question.

Progress should be measured by explanations, implementations, and critical decisions—not by the number of links opened.