probabilistic

Definition

Treating quantities as distributions rather than single values. The payoff is that uncertainty becomes explicit: the model can indicate how sure it is, samples can be drawn instead of committing to one answer, and training can be framed as maximizing likelihood. Note that a probabilistic model is not automatically nondeterministic — varying output across runs comes from sampling at decoding time, and greedy decoding returns the same answer every time.

Explainers using this term

• Build Your Own Diffusion Model — Starting from MNISTDenoising Diffusion Probabilistic Models
• Diffusion Models from the Ground Up — Add Noise, Then Subtract ItDenoising Diffusion Probabilistic Models