https://github.com/philipperemy/neural-probability-dist-sampler

Training a network to sample from any probability distributions such as exponential distribution.
https://github.com/philipperemy/neural-probability-dist-sampler

deep-learning deep-neural-networks probability-distribution sampler statistics tensorflow

Last synced: about 1 month ago
JSON representation

Training a network to sample from any probability distributions such as exponential distribution.

• Host: GitHub
• URL: https://github.com/philipperemy/neural-probability-dist-sampler
• Owner: philipperemy
• Created: 2017-10-15T09:45:00.000Z (over 7 years ago)
• Default Branch: master
• Last Pushed: 2017-10-15T09:57:04.000Z (over 7 years ago)
• Last Synced: 2024-10-29T12:58:05.555Z (3 months ago)
• Topics: deep-learning, deep-neural-networks, probability-distribution, sampler, statistics, tensorflow
• Language: Python
• Size: 2.58 MB
• Stars: 3
• Watchers: 1
• Forks: 2
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# Neural Probability Distribution Sampler

*A end to end differentiable neural network able to sample from any probability distribution, such as the exponential distribution.*

The major requirement is to have a sufficiently larger number of realizations, drawn from the target distribution.

The neural sampler is controlled by only one parameter: the seed. It controls the randomness in the generation. For two identical seeds, the generated realizations should match exactly.

NB: This first implementation restricts the sampling to univariate random variables.



  

  
Normal Distribution -> Exponential Distribution (during training phase)