# numpyro_forecast

> A JAX/NumPyro port of the ideas in Pyro's forecasting module.

## Docs

### API Reference

#### Model building blocks
> Plain model functions that register the train/forecast sites for you.

- [models.Horizon](https://juanitorduz.github.io/numpyro_forecast/reference/models.Horizon.html)
- [models.Transition](https://juanitorduz.github.io/numpyro_forecast/reference/models.Transition.html)
- [models.Advance](https://juanitorduz.github.io/numpyro_forecast/reference/models.Advance.html)
- [models.innovations](https://juanitorduz.github.io/numpyro_forecast/reference/models.innovations.html)
- [models.markov_series](https://juanitorduz.github.io/numpyro_forecast/reference/models.markov_series.html)
- [models.ssoe](https://juanitorduz.github.io/numpyro_forecast/reference/models.ssoe.html)
- [models.SSOEMean](https://juanitorduz.github.io/numpyro_forecast/reference/models.SSOEMean.html)
- [models.SSOEUpdate](https://juanitorduz.github.io/numpyro_forecast/reference/models.SSOEUpdate.html)
- [models.SSOEResult](https://juanitorduz.github.io/numpyro_forecast/reference/models.SSOEResult.html)
- [models.predict](https://juanitorduz.github.io/numpyro_forecast/reference/models.predict.html)
- [models.PlateName](https://juanitorduz.github.io/numpyro_forecast/reference/models.PlateName.html)

#### Vector autoregression
> VAR components that compose with `ssoe` and `markov_series`: conditional mean, step factory, companion matrix, impulse responses.

- [var.var_mean](https://juanitorduz.github.io/numpyro_forecast/reference/var.var_mean.html)
- [var.var_step](https://juanitorduz.github.io/numpyro_forecast/reference/var.var_step.html)
- [var.companion_matrix](https://juanitorduz.github.io/numpyro_forecast/reference/var.companion_matrix.html)
- [var.impulse_response](https://juanitorduz.github.io/numpyro_forecast/reference/var.impulse_response.html)

#### Priors
> Shrinkage prior moments for coefficient arrays.

- [priors.minnesota_prior](https://juanitorduz.github.io/numpyro_forecast/reference/priors.minnesota_prior.html)

#### Distribution surgery
> Time-axis operations on observation distributions, extensible via singledispatch.

- [surgery.shift_loc](https://juanitorduz.github.io/numpyro_forecast/reference/surgery.shift_loc.html)
- [surgery.slice_time](https://juanitorduz.github.io/numpyro_forecast/reference/surgery.slice_time.html)
- [surgery.prefix_condition](https://juanitorduz.github.io/numpyro_forecast/reference/surgery.prefix_condition.html)
- [surgery.register_elementwise](https://juanitorduz.github.io/numpyro_forecast/reference/surgery.register_elementwise.html)

#### Reparameterization
> Time-axis reparameterization of in-sample latents (Haar / DCT), after Pyro's `time_reparam`.

- [reparam.time_reparam](https://juanitorduz.github.io/numpyro_forecast/reference/reparam.time_reparam.html)
- [reparam.TimeTransform](https://juanitorduz.github.io/numpyro_forecast/reference/reparam.TimeTransform.html)

#### Producing draws
> Drawing posterior samples and generating forecasts and in-sample predictions.

- [predictive.draw_posterior](https://juanitorduz.github.io/numpyro_forecast/reference/predictive.draw_posterior.html)
- [predictive.forecast](https://juanitorduz.github.io/numpyro_forecast/reference/predictive.forecast.html)
- [predictive.predict_in_sample](https://juanitorduz.github.io/numpyro_forecast/reference/predictive.predict_in_sample.html)

#### Backtesting & evaluation
> Rolling-window backtesting and forecast metrics.

- [evaluate.backtest](https://juanitorduz.github.io/numpyro_forecast/reference/evaluate.backtest.html)
- [evaluate.backtest_vectorized](https://juanitorduz.github.io/numpyro_forecast/reference/evaluate.backtest_vectorized.html)
- [evaluate.BacktestResult](https://juanitorduz.github.io/numpyro_forecast/reference/evaluate.BacktestResult.html)
- [evaluate.VectorizedBacktestResult](https://juanitorduz.github.io/numpyro_forecast/reference/evaluate.VectorizedBacktestResult.html)
- [evaluate.WindowType](https://juanitorduz.github.io/numpyro_forecast/reference/evaluate.WindowType.html)
- [evaluate.DEFAULT_METRICS](https://juanitorduz.github.io/numpyro_forecast/reference/evaluate.DEFAULT_METRICS.html)
- [evaluate.evaluate_forecast](https://juanitorduz.github.io/numpyro_forecast/reference/evaluate.evaluate_forecast.html)
- [evaluate.results_to_dataframe](https://juanitorduz.github.io/numpyro_forecast/reference/evaluate.results_to_dataframe.html)
- [evaluate.eval_crps](https://juanitorduz.github.io/numpyro_forecast/reference/evaluate.eval_crps.html)
- [evaluate.eval_mae](https://juanitorduz.github.io/numpyro_forecast/reference/evaluate.eval_mae.html)
- [evaluate.eval_rmse](https://juanitorduz.github.io/numpyro_forecast/reference/evaluate.eval_rmse.html)
- [evaluate.eval_coverage](https://juanitorduz.github.io/numpyro_forecast/reference/evaluate.eval_coverage.html)
- [metrics.crps_empirical](https://juanitorduz.github.io/numpyro_forecast/reference/metrics.crps_empirical.html)
- [metrics.eval_pinball](https://juanitorduz.github.io/numpyro_forecast/reference/metrics.eval_pinball.html)
- [metrics.eval_interval_score](https://juanitorduz.github.io/numpyro_forecast/reference/metrics.eval_interval_score.html)
- [metrics.make_mase](https://juanitorduz.github.io/numpyro_forecast/reference/metrics.make_mase.html)

#### ArviZ export
> Convert posteriors into ArviZ-schema xarray DataTrees for diagnostics and plotting.

- [convert.to_datatree](https://juanitorduz.github.io/numpyro_forecast/reference/convert.to_datatree.html)
- [convert.add_forecast_groups](https://juanitorduz.github.io/numpyro_forecast/reference/convert.add_forecast_groups.html)
- [convert.predictions_to_datatree](https://juanitorduz.github.io/numpyro_forecast/reference/convert.predictions_to_datatree.html)

#### Extensions (contrib)
> Optional backends behind pyproject extras (never imported by default).

- [contrib.blackjax.BlackjaxNUTSKernel](https://juanitorduz.github.io/numpyro_forecast/reference/contrib.blackjax.BlackjaxNUTSKernel.html)
- [contrib.blackjax.BlackjaxMCLMCKernel](https://juanitorduz.github.io/numpyro_forecast/reference/contrib.blackjax.BlackjaxMCLMCKernel.html)
- [contrib.blackjax.BlackjaxCustomKernel](https://juanitorduz.github.io/numpyro_forecast/reference/contrib.blackjax.BlackjaxCustomKernel.html)
- [contrib.blackjax.PathfinderFit](https://juanitorduz.github.io/numpyro_forecast/reference/contrib.blackjax.PathfinderFit.html)
- [contrib.blackjax.fit_pathfinder](https://juanitorduz.github.io/numpyro_forecast/reference/contrib.blackjax.fit_pathfinder.html)
- [contrib.blackjax.pathfinder_samples](https://juanitorduz.github.io/numpyro_forecast/reference/contrib.blackjax.pathfinder_samples.html)
- [contrib.blackjax.MultiPathfinderFit](https://juanitorduz.github.io/numpyro_forecast/reference/contrib.blackjax.MultiPathfinderFit.html)
- [contrib.blackjax.fit_multipathfinder](https://juanitorduz.github.io/numpyro_forecast/reference/contrib.blackjax.fit_multipathfinder.html)
- [contrib.blackjax.multipathfinder_samples](https://juanitorduz.github.io/numpyro_forecast/reference/contrib.blackjax.multipathfinder_samples.html)

#### Typing
> Public type contracts.

- [typing.ForecastModel](https://juanitorduz.github.io/numpyro_forecast/reference/typing.ForecastModel.html)
- [typing.ForecastFn](https://juanitorduz.github.io/numpyro_forecast/reference/typing.ForecastFn.html)
- [typing.Guide](https://juanitorduz.github.io/numpyro_forecast/reference/typing.Guide.html)
- [typing.InSampleFn](https://juanitorduz.github.io/numpyro_forecast/reference/typing.InSampleFn.html)
- [typing.Metric](https://juanitorduz.github.io/numpyro_forecast/reference/typing.Metric.html)
- [typing.ModelFactory](https://juanitorduz.github.io/numpyro_forecast/reference/typing.ModelFactory.html)
- [typing.Array](https://juanitorduz.github.io/numpyro_forecast/reference/typing.Array.html)
- [typing.BlackjaxBuildFn](https://juanitorduz.github.io/numpyro_forecast/reference/typing.BlackjaxBuildFn.html)

#### Autocorrelation
> Batched autocorrelation and partial autocorrelation diagnostics.

- [acf.acf](https://juanitorduz.github.io/numpyro_forecast/reference/acf.acf.html)
- [acf.pacf](https://juanitorduz.github.io/numpyro_forecast/reference/acf.pacf.html)

#### Seasonal features
> Fourier design matrices and seasonal tiling.

- [features.fourier_features](https://juanitorduz.github.io/numpyro_forecast/reference/features.fourier_features.html)
- [features.periodic_repeat](https://juanitorduz.github.io/numpyro_forecast/reference/features.periodic_repeat.html)

#### Array helpers
> Time-axis array shaping for the train/forecast split.

- [arrays.zero_data_like](https://juanitorduz.github.io/numpyro_forecast/reference/arrays.zero_data_like.html)
- [arrays.concat_future](https://juanitorduz.github.io/numpyro_forecast/reference/arrays.concat_future.html)
- [arrays.pad_future](https://juanitorduz.github.io/numpyro_forecast/reference/arrays.pad_future.html)

#### Datasets
> Example datasets used in the tutorials.

- [datasets.load_bart_weekly](https://juanitorduz.github.io/numpyro_forecast/reference/datasets.load_bart_weekly.html)
- [datasets.load_bart_hierarchical](https://juanitorduz.github.io/numpyro_forecast/reference/datasets.load_bart_hierarchical.html)
- [datasets.load_victoria_electricity](https://juanitorduz.github.io/numpyro_forecast/reference/datasets.load_victoria_electricity.html)
- [datasets.bart_available](https://juanitorduz.github.io/numpyro_forecast/reference/datasets.bart_available.html)
- [datasets.load_m5](https://juanitorduz.github.io/numpyro_forecast/reference/datasets.load_m5.html)
- [datasets.M5Data](https://juanitorduz.github.io/numpyro_forecast/reference/datasets.M5Data.html)

#### Optional dependencies
> Lazy imports behind pyproject extras.

- [optional.require](https://juanitorduz.github.io/numpyro_forecast/reference/optional.require.html)

#### Exceptions
> Package exception hierarchy raised at validation boundaries.

- [exceptions.NumpyroForecastError](https://juanitorduz.github.io/numpyro_forecast/reference/exceptions.NumpyroForecastError.html)
- [exceptions.BacktestWindowError](https://juanitorduz.github.io/numpyro_forecast/reference/exceptions.BacktestWindowError.html)
- [exceptions.VectorizedMetricError](https://juanitorduz.github.io/numpyro_forecast/reference/exceptions.VectorizedMetricError.html)
- [exceptions.KernelConfigError](https://juanitorduz.github.io/numpyro_forecast/reference/exceptions.KernelConfigError.html)
- [exceptions.CovariateDimsError](https://juanitorduz.github.io/numpyro_forecast/reference/exceptions.CovariateDimsError.html)
- [exceptions.MVNLayoutError](https://juanitorduz.github.io/numpyro_forecast/reference/exceptions.MVNLayoutError.html)
- [exceptions.DeviceMemoryError](https://juanitorduz.github.io/numpyro_forecast/reference/exceptions.DeviceMemoryError.html)
- [exceptions.HostMemoryKindError](https://juanitorduz.github.io/numpyro_forecast/reference/exceptions.HostMemoryKindError.html)
- [exceptions.DevicePlatformError](https://juanitorduz.github.io/numpyro_forecast/reference/exceptions.DevicePlatformError.html)