0
$\begingroup$

Is there way to generate Exponential(lambda) distributed samples via a reparameterization trick?

As in: Reparameterization trick for gamma distribution

And also: How does the reparameterization trick for VAEs work and why is it important?

My motivation is to compare Exponential priors with Normal priors, in a neural network, having the latter solved as per the second link.

Though I have no experience in this field (and a general understanding), I am coding a project in MXNet where I compare different architectures in terms of predictive accuracy on IHDP data.

$\endgroup$
2
$\begingroup$

If you can generate an exponential rv. with rate 1 (call this X), then you can generate an exponential random variable with rate lambda by multiplying X with 1/lambda.

So, if you want to use an exponential function as a posterior, you can use the re-parametrization trick in variational autoencoders by only generating exponential rv. with rate 1.

| cite | improve this answer | |
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.