State-Centered Temporal Processes#828
Open
cdc-mitzimorris wants to merge 81 commits into
Open
Conversation
…e time 0) (#827) * bug fix and unit tests * Potential fix for pull request finding Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com> * update unit test to match code * revert changes, apply simpler fix --------- Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com>
for more information, see https://pre-commit.ci
…v/PyRenew into mem_810_centered_parameterization
for more information, see https://pre-commit.ci
…v/PyRenew into mem_810_centered_parameterization
for more information, see https://pre-commit.ci
Collaborator
Author
|
@dylanhmorris @sbidari ready for code review |
…v/PyRenew into mem_810_centered_parameterization
Collaborator
|
Linking epiforecasts/EpiNow2#1396 because it seems to cover a similar reparametrisation |
Collaborator
Author
thanks! good to know there's good precedent. |
Collaborator
Author
|
To simplify this PR, I propose splitting out the The code added/ modified by this PR, exclusive of benchmarks code is:
API-change subtotal: ~1,673 added / ~182 removed. Other changes also picked up on this branch (not in the PR description)
Other-changes subtotal: ~40 added / ~104 removed. Benchmarking component adds roughly 3K lines of code. (oof!)
Totals: 13 source files (~2,346 lines) + 718 lines of tests. The benchmark framework is self-contained under |
Collaborator
Author
|
@dylanhmorris and @sbidari - ready for review. |
Comment on lines
+681
to
+682
| class TestStateCenteredRandomWalk: | ||
| """State-centered RandomWalk samples the state path directly via GaussianRandomWalk.""" |
Comment on lines
+16
to
+28
| class StateRandomWalk(Distribution): | ||
| r""" | ||
| State-centered random-walk prior on a post-initial state path. | ||
|
|
||
| Given a deterministic initial state $x_0$ = ``initial_loc``: | ||
|
|
||
| $$ | ||
| x_t \sim \mathrm{Normal}(x_{t-1}, \sigma), \quad t = 1, \dots, T | ||
| $$ | ||
|
|
||
| The sampled value is the post-initial path | ||
| $[x_1, x_2, \ldots, x_{\mathrm{num\_steps}}]$ of length ``num_steps``. | ||
| """ |
Comment on lines
+283
to
+287
| Multiplies the finite entries of ``predicted`` by the weekday | ||
| cycle anchored at ``first_day_dow``. ``NaN`` entries (the | ||
| delay-tail at the start of the shared time axis) are preserved | ||
| through the JAX "double-where" idiom: the inner product is | ||
| evaluated against a NaN-free surrogate so its backward |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
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.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Added state-centered parameterizations for all three temporal-process
classes in
pyrenew.latent:AR1— stationary AR(1) on log-Rt levelsDifferencedAR1— AR(1) on first differences of log-Rt (the productionprocess)
RandomWalk— unconstrained drift on log-RtEach class now takes a constructor argument
parameterization: Literal["innovation", "state"], defaulting to"innovation"to preserve current behavior. Setting"state"switchesthe internal sampling from standardized increments to the latent state
path directly.
The state-centered variants are implemented via:
RandomWalk: NumPyro's built-indist.GaussianRandomWalk, shiftedby the initial value.
AR1andDifferencedAR1: two new custom NumPyroDistributionsubclasses (
StateAR1,StateDifferencedAR1) inpyrenew/latent/state_centered_distributions.py. Both have vectorizedlog_probusing slice arithmetic (no scan during MCMC) andlax.scan-basedsample(only called for prior/posterior predictive,not on the MCMC gradient path).
Both parameterizations encode the same prior distribution over the
state path. They differ only in sampler geometry — which latent
variables HMC sees and operates on.
Code added
pyrenew/latent/state_centered_distributions.pyStateAR1,StateDifferencedAR1pyrenew/latent/temporal_processes.pyparameterizationflag on all three classes;_prepare_initial_valuehelpertest/test_temporal_processes.pytest/test_helpers.pyfixed_ar1_state,fixed_differenced_ar1_statefactoriestest/integration/conftest.pyhe_model_state_centered,he_weekly_rt_model_state_centered,he_weekly_model_state_centeredfixturestest/integration/test_population_infections_he_state_centered.pytest/integration/test_population_infections_he_weekly_rt_state_centered.pyWeeklyTemporalProcess_typos.tomlreparametrized_params(NumPyro upstream attribute name)