diffuusiopohjaiset
Diffuusiopohjaiset, often translated as diffusion-based, refers to a category of models that operate on the principle of diffusion. These models are characterized by a process that gradually adds noise to data until it becomes indistinguishable from random noise. Subsequently, a reverse process is employed to denoise the data, ultimately reconstructing the original information or generating new data that resembles the training distribution.
In machine learning, diffusion models have gained significant traction, particularly in the field of generative modeling.
The mathematical framework behind diffusion models typically involves stochastic differential equations or Markov chains. These models