Unconditional diffusion of hit patterns.
- PyTorch with UNet (https://arxiv.org/pdf/1505.04597.pdf)
- Training and testing based on simulation for: CERN 18
- run 403, 7 GeV/c pions @ 30 degree polar angle
Unconditional training with UNet. Network learns to to determine amount of noise in the input:
Unconditional generation with UNet:
Training samples for 5k pions at different polar angle:
Unconditional diffusion with 50 steps:
Next step: adding conditions: