About
TheoremPath is a structured theory library for machine learning, statistics, and mathematics.
Every topic states its prerequisites. Every theorem lists its assumptions. Every major claim cites a source. The site is built for readers who want more than intuition alone: precise definitions, explicit conditions, and a clear path from foundations to modern ML.
What's here
- 491 topic pages spanning probability, statistics, optimization, learning theory, deep learning, and reinforcement learning.
- A prerequisite graph linking topics back to their foundations. The Atlas lets you trace any topic back to its axioms by following the edges in reverse.
- Interactive demos for bias-variance, softmax temperature, dropout, optimization, and likelihood geometry.
- Guided learning paths for different backgrounds (math to ML, engineer to theory, stats to deep learning, research interview prep).
- A diagnostic that surfaces likely gaps in your foundations and points you at the next topics to read.
How pages are written
Theorems state assumptions explicitly. Bounds are stated with constants when the constants matter. Every theorem page includes failure modes: the conditions under which the result breaks.
Topics are organized by foundational depth, importance, and recency, so readers can separate timeless material from current or frontier ideas.
References
Foundational topics cite canonical textbooks with chapter-level references. Topics covering post-2015 deep learning also cite the primary papers directly.
Primary papers (examples):
- He, Zhang, Ren, Sun. Deep Residual Learning for Image Recognition (2015).
- Kingma, Ba. Adam: A Method for Stochastic Optimization (2014).
- Mnih et al. Human-level control through deep reinforcement learning (Nature, 2015).
- Silver et al. Mastering the game of Go without human knowledge (Nature, 2017).
- Vaswani et al. Attention Is All You Need (2017).
- Ho, Jain, Abbeel. Denoising Diffusion Probabilistic Models (2020).
- Kaplan et al. Scaling Laws for Neural Language Models (2020).
- Radford et al. Learning Transferable Visual Models From Natural Language Supervision (CLIP) (2021).
- Hoffmann et al. Training Compute-Optimal Large Language Models (Chinchilla) (2022).
- Ouyang et al. Training language models to follow instructions with human feedback (InstructGPT) (2022).
Canonical textbook shelf, cited throughout:
- Goodfellow, Bengio, Courville. Deep Learning (2016).
- Hastie, Tibshirani, Friedman. The Elements of Statistical Learning (2009).
- Russell and Norvig. Artificial Intelligence: A Modern Approach, 4th ed. (2020).
- Shalev-Shwartz and Ben-David. Understanding Machine Learning (2014).
- Mohri, Rostamizadeh, Talwalkar. Foundations of Machine Learning (2018).
- Vershynin. High-Dimensional Probability (2018).
- Wainwright. High-Dimensional Statistics (2019).
- Boyd and Vandenberghe. Convex Optimization (2004).
- Durrett. Probability: Theory and Examples (2019).
- Casella and Berger. Statistical Inference (2002).
- Sutton and Barto. Reinforcement Learning: An Introduction (2018).
- Cover and Thomas. Elements of Information Theory (2006).
Stack
TypeScript, Next.js, Tailwind, MDX, KaTeX. Content is file-based, version-controlled, and statically generated.
Maintained by Robby Sneiderman.