This isn’t a thing per se.
But I want a landing page for a few different ways to think about hidden variables.
If the hidden varaibles are random, I might think of hierarchical models,
latent factors, deep networks, informative sampling, censored data…
If the hidden variables are deterministic I might think of algorithmic complexity.
Matrix factorisation to approximate latent factors and their interaction in recommender systems.
- Koren, Y., Bell, R., & Volinsky, C. (2009) Matrix Factorization Techniques for Recommender Systems. Computer, 42(8), 30–37. DOI.