Prior Machine Learning
noun (often used adjectivally in compounds, e.g., prior probability)
Definition: In Bayesian inference, a prior (or prior distribution) is a probability distribution that represents beliefs about a quantity or parameter before new data are observed [Bakanach 2023/2024].
Example in context: “A prior distribution is chosen arbitrarily by a model designer, such as a unit Normal distribution N (0, 1).” [Asai et al. 2022]
“The prior is such that both segments are equally likely, and that Pr (cid:2)g2|S2(cid:3) > 3 Pr (cid:2)g1|S2(cid:3) > 0.” [Gradwohl, Tennenholtz 2023]
Synonym: prior distribution (when distribution is explicit)
Related terms: prior probability, likelihood, posterior, Bayesian updating