Closed FindTheAnswer closed 3 years ago
The deterministic model predicts a single result, while the stochastic model generates a prediction of distribution.
To convert a deterministic model to a stochastic version, one simple approach would be to introduce Gaussian noise when generating predictions and sample from the distribution.
The deterministic model predicts a single result, while the stochastic model generates a prediction of distribution.
To convert a deterministic model to a stochastic version, one simple approach would be to introduce Gaussian noise when generating predictions and sample from the distribution.
Many thanks! And to convert a stochastic version to a deterministic version, I should throw away Gaussian noise when training and generating predictions or just when generating predictions?
You are most welcome.
Yes, you should train the model without Gaussian noise.
However, for this repo, we only provide the stochastic version of STAR. The hyperparameters of the deterministic version and the stochastic version can be quite different.
You are most welcome.
Yes, you should train the model without Gaussian noise.
However, for this repo, we only provide the stochastic version of STAR. The hyperparameters of the deterministic version and the stochastic version can be quite different.
Have you considered releasing the deterministic version?
We currently have no plan to release it since I'm quite occupied with other stuff.
However, we may try to release it once I have more free time to deal with it.
We currently have no plan to release it since I'm quite occupied with other stuff.
However, we may try to release it once I have more free time to deal with it.
Thanks again!
Hi, what is the difference between Deterministic model and Stochastic model? And what should I do to change the Deterministic version to Stochastic version?