생성 AI
1.Forward Diffusion Process
$$q(𝐱{1:T}|𝐱0)=∏{t=1}^Tq(𝐱t|𝐱_{t-1})$$ $$q(𝐱t|𝐱{t-1})=𝒩(𝐱t;\sqrt{1-\betat}𝐱{t-1},\betat𝐈)$$ We can sample at any arbitary time step $$t$$ by using reparameterization trick Let $$\alphat =...
2025년 1월 23일
$$q(𝐱{1:T}|𝐱0)=∏{t=1}^Tq(𝐱t|𝐱_{t-1})$$ $$q(𝐱t|𝐱{t-1})=𝒩(𝐱t;\sqrt{1-\betat}𝐱{t-1},\betat𝐈)$$ We can sample at any arbitary time step $$t$$ by using reparameterization trick Let $$\alphat =...