[HN Gopher] What Are Diffusion Models?
       ___________________________________________________________________
        
       What Are Diffusion Models?
        
       Author : headalgorithm
       Score  : 107 points
       Date   : 2021-07-12 10:49 UTC (12 hours ago)
        
 (HTM) web link (lilianweng.github.io)
 (TXT) w3m dump (lilianweng.github.io)
        
       | heydenberk wrote:
       | I can't recommend this blog highly enough. Just about every post
       | provides a deeply detailed, understandable overview on a subject
       | at the frontier of ML research
        
         | btowngar wrote:
         | Completely agreed, a fantastic blog.
        
           | stevofolife wrote:
           | I also find Eugene Yan's less technical blog to be similarly
           | fantastic. https://eugeneyan.com/
        
       | MPSimmons wrote:
       | Not being completely fluent in the ML space, I'm not completely
       | clear on the applicability of this technique.
       | 
       | Broadly, is this useful for determining the already-diffused
       | pattern of data and determining the original inputs, or
       | determining the diffused output without needing to iterate fully
       | and produce the result manually, or both, or am I completely off?
        
         | eutectic wrote:
         | They are a way of training generative models similar to GANs or
         | autoencoders.
         | 
         | My understanding is that if you train an autoencoder with a
         | gaussian likelihood then you will tend to get fuzzy samples,
         | but using an iterative process where each step is a gaussian
         | conditioned on the previous step can give you nicer samples.
        
       | nharada wrote:
       | Thanks for the link and the blog post. Diffusion models have
       | definitely been making waves over the last year or two and I've
       | been slacking on really digging in.
        
       ___________________________________________________________________
       (page generated 2021-07-12 23:02 UTC)