arjit_jain

7 ms·

1. 1.▲
   by arjit_jain 6y ago

   Great catch! Importance splitting is indeed a useful simulation technique for low probability events. The application of splitting based methods on bayesian models for extreme sample generation can be an interesting future work.
2. 2.▲
   by arjit_jain 6y ago

   Existing GAN based approaches excel at generating realistic samples, but seek to generate typical samples, rather than extreme samples. We propose ExGAN to generate realistic and extreme samples. ExGAN allows the user to specify both the de
3. 3.▲
   Show HN: Adversarial Generation of Extreme Samples(arxiv.org)

   35 points by arjit_jain 6y ago|4 comments