← All tracks

JAX Stochasticity

Distributions, reparameterization, sampling techniques, MCMC, gradient estimators. Randomness as a controllable resource.

0 / 25 solved Continue →
  1. 1. ○ Uniform Sampling
  2. 2. ○ Normal Sampling with mean and std
  3. 3. ○ Bernoulli Mask Sampling
  4. 4. ○ Categorical Sampling
  5. 5. ○ Reparameterization Trick: Gaussian
  6. 6. ○ Gumbel-Softmax
  7. 7. ○ Dirichlet Sampling
  8. 8. ○ Temperature Scaling
  9. 9. ○ Top-k Logit Masking
  10. 10. ○ Nucleus (Top-p) Masking
  11. 11. ○ Gumbel Argmax (Categorical via Trick)
  12. 12. ○ Metropolis-Hastings Step
  13. 13. ○ Log Acceptance Ratio
  14. 14. ○ HMC Leapfrog Step
  15. 15. ○ REINFORCE Gradient Estimator
  16. 16. ○ Reparameterization Gradient
  17. 17. ○ REINFORCE with Baseline
  18. 18. ○ Batched Sampling via vmap
  19. 19. ○ Random Walk via lax.scan
  20. 20. ○ Importance Sampling
  21. 21. ○ Beta Distribution Sampling
  22. 22. ○ Multivariate Normal Sampling
  23. 23. ○ Random Permutation
  24. 24. ○ Poisson Sampling
  25. 25. ○ ELBO for Gaussian VI