fix - in chapter 4, correct the bspline model section (changed in final copy of the book) #6
Conversation
…al copy of the book)
|
Thanks for the update, @ksachdeva! The diff is so large due to
I think the reason is the parameters have not been optimized by Adam optimizer yet. You would need to observe the loss and change learning rate, the number of iterations accordingly. :) After jax 0.1.74, there is BFGS optimizer which does not require tuning learning rate but we haven't included it in NumPyro yet. I think SR author also uses BFGS optimizer for his book. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
In the final edition, for the BSpline model, the
doycolumn is used (instead oftemp). The loc and scale in the model are also very different from December 2019 draftIn Code 4.76, you will notice a comment regarding explicit state initialization of
ato 100. Without this model fails to converge. I had tried to increase the sample size but it was in vain. Witha's initial state set around 100, I am able to correctly reproduce the figures from the book.