danesabo/Obsidian
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

vault backup: 2024-10-28 09:09:26

Browse Source
This commit is contained in:
Dane Sabo 2024-10-28 09:09:26 -04:00
parent 00f099b607
commit a779ae4081
1 changed files with 10 additions and 7 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

17
4 Qualifying Exam/2 Writing/3. QE Research Approach.md
View File

@ -29,10 +29,13 @@ Something to justify, why diffusion model as opposed to other generative AI
6. Diffusion model uses two processes 6. Diffusion model uses two processes
7. a forward process that introduces noise into an 'input' 7. a forward process that introduces noise into an 'input'
8. this forward process does this using several small steps of noise 8. this forward process does this using several small steps of noise
9. This noise is a gaussian noise 9. These are called timesteps
10. over time this forward process degrades the input until it is unrecognizable 10. This noise is a gaussian noise
11. This takes several several iterations, but is dependent on the amount of noise in each step 11. over time this forward process degrades the input until it is unrecognizable
12. The markov chain that this creates is also gaussian at every step, until the noise at the end of the day is some gaussian distribution. Usually mean 0 std. dev beta 12. This takes several several iterations, but is dependent on the amount of noise in each step
13. A reverse process that tries to remove the noise 13. The markov chain that this creates is also gaussian at every step, until the noise at the end of the day is some gaussian distribution. Usually mean 0 std. dev beta
14. But if we destroy the input how can we do this? 14. A reverse process that tries to remove the noise
15. Well we train a neural network as a denoiser. 15. But if we destroy the input how can we do this?
16. Well we train a neural network as a denoiser.
17. Because the diffusion model forward steps are small and gaussian, we can know the reverse step is also a gaussian distribution.
18. So for our neural network, what we're trying to learn is the mean and standard deviation of the reverse steps for a given timestep.