Merge remote-tracking branch 'upstream/master'

Author: fonnesbeck

.pre-commit-config.yaml

@@ -1,17 +1,17 @@
 repos:
 - repo: https://github.com/pre-commit/pre-commit-hooks
-  rev: v3.2.0
+  rev: v3.3.0
   hooks:
     -   id: end-of-file-fixer
     -   id: check-toml
 - repo: https://github.com/nbQA-dev/nbQA
-  rev: 0.3.2
+  rev: 0.3.5
   hooks:
     - id: nbqa-black
     - id: nbqa-isort
     - id: nbqa-pyupgrade
 - repo: https://github.com/asottile/pyupgrade
-  rev: v2.7.2
+  rev: v2.7.3
   hooks:
     - id: pyupgrade
       args: [--py36-plus]
@@ -22,7 +22,14 @@ repos:
 - repo: local
   hooks:
     - id: watermark
-      name: Check notebooks have watermark
+      name: Check notebooks have watermark (see Jupyter style guide from PyMC3 Uncyclo)
       types: [jupyter]
-      entry: python scripts/check_watermark.py
+      entry: '%load_ext watermark.*%watermark -n -u -v -iv -w'
+      language: pygrep
+      args: [--negate, --multiline]
+      minimum_pre_commit_version: 2.8.0
+    - id: check-toc
+      name: Check all notebooks appear in table of contents
+      types: [jupyter]
+      entry: python scripts/check_toc_is_complete.py
       language: python

README.rst

@@ -15,6 +15,18 @@ Check out the `getting started guide <http://docs.pymc.io/notebooks/getting_star
 using Binder!
 For questions on PyMC3, head on over to our `PyMC Discourse <https://discourse.pymc.io/>`__ forum.
 
+The future of PyMC3 & Theano
+============================
+
+There have been many questions and uncertainty around the future of PyMC3 since Theano
+stopped getting developed by the original authors, and we started experiments with PyMC4.
+
+We are happy to announce that PyMC3 on Theano (which we are `developing further <https://github.com/pymc-devs/Theano-PyMC>`__) 
+with a new JAX backend is the future. PyMC4 will not be developed further. 
+
+See the `full announcement <https://pymc-devs.medium.com/the-future-of-pymc3-or-theano-is-dead-long-live-theano-d8005f8a0e9b>`__
+for more details. 
+
 Features
 ========
 

RELEASE-NOTES.md

@@ -16,11 +16,13 @@
 - `sample_posterior_predictive_w` can now feed on `xarray.Dataset` - e.g. from `InferenceData.posterior`. (see [#4042](https://github.com/pymc-devs/pymc3/pull/4042))
 - Added `pymc3.gp.cov.Circular` kernel for Gaussian Processes on circular domains, e.g. the unit circle (see [#4082](https://github.com/pymc-devs/pymc3/pull/4082)).
 - Add MLDA, a new stepper for multilevel sampling. MLDA can be used when a hierarchy of approximate posteriors of varying accuracy is available, offering improved sampling efficiency especially in high-dimensional problems and/or where gradients are not available (see [#3926](https://github.com/pymc-devs/pymc3/pull/3926))
-- Change SMC metropolis kernel to independent metropolis kernel [#4115](https://github.com/pymc-devs/pymc3/pull/3926))
+- Change SMC metropolis kernel to independent metropolis kernel [#4115](https://github.com/pymc-devs/pymc3/pull/4115))
 - Add alternative parametrization to NegativeBinomial distribution in terms of n and p (see [#4126](https://github.com/pymc-devs/pymc3/issues/4126))
+- Add Bayesian Additive Regression Trees (BARTs) [#4183](https://github.com/pymc-devs/pymc3/pull/4183))
 - Added a new `MixtureSameFamily` distribution to handle mixtures of arbitrary dimensions in vectorized form (see [#4185](https://github.com/pymc-devs/pymc3/issues/4185)).
 
 
+
 ## PyMC3 3.9.3 (11 August 2020)
 
 ### Maintenance

docs/source/notebooks/GLM-hierarchical.ipynb

@@ -36,7 +36,7 @@
     "Gelman et al.'s (2007) radon dataset is a classic for hierarchical modeling. In this dataset the amount of the radioactive gas radon has been measured among different households in all counties of several states. Radon gas is known to be the highest cause of lung cancer in non-smokers. It is believed to be more strongly present in households containing a basement and to differ in amount present among types of soil.\n",
     "Here we'll investigate this differences and try to make predictions of radonlevels in different counties based on the county itself and the presence of a basement. In this example we'll look at Minnesota, a state that contains 85 counties in which different measurements are taken, ranging from 2 to 116 measurements per county. \n",
     "\n",
-    "![radon](http://www.fix-your-radon.com/images/how_radon_enters.jpg)"
+    "![radon](https://upload.wikimedia.org/wikipedia/commons/b/b9/CNX_Chem_21_06_RadonExpos.png)"
    ]
   },
   {

docs/source/notebooks/api_quickstart.ipynb

@@ -998,7 +998,7 @@
     "\n",
     "The main entry point to MCMC sampling algorithms is via the `pm.sample()` function. By default, this function tries to auto-assign the right sampler(s) and auto-initialize if you don't pass anything.\n",
     "\n",
-    "With PyMC3 version >=3.9 the `return_inferencedata=True` kwarg makes the `sample` function return an `arviz.InferenceData` object instead of a `MultiTrace`. `InferenceData` has many advantages, compared to a `MultiTrace`: For example it can be saved/loaded from a file, and can also carry additional (meta)data such as date/version, or posterior predictive distributions. Take a look at the [ArviZ Quickstart](https://arviz-devs.github.io/arviz/notebooks/Introduction.html) to learn more."
+    "With PyMC3 version >=3.9 the `return_inferencedata=True` kwarg makes the `sample` function return an `arviz.InferenceData` object instead of a `MultiTrace`. `InferenceData` has many advantages, compared to a `MultiTrace`: For example it can be saved/loaded from a file, and can also carry additional (meta)data such as date/version, or posterior predictive distributions. Take a look at the [ArviZ Quickstart](https://arviz-devs.github.io/arviz/getting_started/Introduction.html) to learn more."
    ]
   },
   {
@@ -2325,7 +2325,7 @@
    "name": "python",
    "nbconvert_exporter": "python",
    "pygments_lexer": "ipython3",
-   "version": "3.7.7"
+   "version": "3.6.8"
   }
  },
  "nbformat": 4,

docs/source/notebooks/blackbox_external_likelihood.ipynb

@@ -51,7 +51,7 @@
     "import theano\n",
     "import theano.tensor as tt\n",
     "\n",
-    "print(\"Running on PyMC3 v{}\".format(pm.__version__))"
+    "print(f\"Running on PyMC3 v{pm.__version__}\")"
    ]
   },
   {
@@ -697,7 +697,7 @@
     "            color=colors[i],\n",
     "            hist_kwargs={\"density\": True},\n",
     "            **hist2dkwargs,\n",
-    "            truths=[mtrue, ctrue]\n",
+    "            truths=[mtrue, ctrue],\n",
     "        )\n",
     "    else:\n",
     "        corner.corner(\n",
@@ -729,15 +729,15 @@
     "test_grad_op_func = theano.function([var], test_grad_op(var))\n",
     "grad_vals = test_grad_op_func([mtrue, ctrue])\n",
     "\n",
-    "print('Gradient returned by \"LogLikeGrad\": {}'.format(grad_vals))\n",
+    "print(f'Gradient returned by \"LogLikeGrad\": {grad_vals}')\n",
     "\n",
     "# test the gradient called through LogLikeWithGrad\n",
     "test_gradded_op = LogLikeWithGrad(my_loglike, data, x, sigma)\n",
     "test_gradded_op_grad = tt.grad(test_gradded_op(var), var)\n",
     "test_gradded_op_grad_func = theano.function([var], test_gradded_op_grad)\n",
     "grad_vals_2 = test_gradded_op_grad_func([mtrue, ctrue])\n",
     "\n",
-    "print('Gradient returned by \"LogLikeWithGrad\": {}'.format(grad_vals_2))\n",
+    "print(f'Gradient returned by \"LogLikeWithGrad\": {grad_vals_2}')\n",
     "\n",
     "# test the gradient that PyMC3 uses for the Normal log likelihood\n",
     "test_model = pm.Model()\n",
@@ -751,7 +751,7 @@
     "    gradfunc.set_extra_values({\"m_interval__\": mtrue, \"c_interval__\": ctrue})\n",
     "    grad_vals_pymc3 = gradfunc(np.array([mtrue, ctrue]))[1]  # get dlogp values\n",
     "\n",
-    "print('Gradient returned by PyMC3 \"Normal\" distribution: {}'.format(grad_vals_pymc3))"
+    "print(f'Gradient returned by PyMC3 \"Normal\" distribution: {grad_vals_pymc3}')"
    ]
   },
   {

docs/source/notebooks/table_of_contents_examples.js

@@ -34,6 +34,7 @@ Gallery.contents = {
     "gaussian_process": "Gaussian Processes",
     "conditional-autoregressive-model": "Gaussian Processes",
     "log-gaussian-cox-process":"Gaussian Processes",
+    "GP-Circular": "Gaussian Processes",
     "dependent_density_regression": "Mixture Models",
     "dp_mix": "Mixture Models",
     "gaussian-mixture-model-advi": "Mixture Models",
@@ -59,6 +60,7 @@ Gallery.contents = {
     "ODE_with_manual_gradients": "Inference in ODE models",
     "ODE_API_introduction": "Inference in ODE models",
     "probabilistic_matrix_factorization": "Case Studies",
+    "MLDA_introduction": "MCMC",
     "MLDA_simple_linear_regression": "MCMC",
     "MLDA_gravity_surveying": "MCMC",
     "MLDA_variance_reduction_linear_regression": "MCMC"

docs/source/notebooks/table_of_contents_tutorials.js

@@ -6,13 +6,15 @@ Gallery.contents = {
     "Gaussian_Processes.rst": "Basics",
     "data_container": "Basics",
     "sampling_compound_step": "Deep dives",
+    "sampling_callback": "Deep dives",
     "sampler-stats": "Deep dives",
     "Diagnosing_biased_Inference_with_Divergences": "Deep dives",
     "Advanced_usage_of_Theano_in_PyMC3.rst": "Deep dives",
     "getting_started": "Deep dives",
     "ODE_API_shapes_and_benchmarking": "Deep dives",
     "DEMetropolisZ_EfficiencyComparison": "Deep dives",
     "DEMetropolisZ_tune_drop_fraction": "Deep dives",
+    "factor_analysis": "Deep dives",
     "blackbox_external_likelihood": "How-To",
     "profiling": "How-To",
     "howto_debugging": "How-To",

environment-dev.yml

@@ -21,6 +21,7 @@ dependencies:
   - nose>=1.3
   - nose-parameterized>=0.6
   - numpydoc>=0.9
+  - pre-commit>=2.8.0
   - pycodestyle>=2.3
   - pyflakes>=1.5
   - pylint>=1.7

pymc3/distributions/__init__.py

@@ -99,8 +99,11 @@
 from .timeseries import MvGaussianRandomWalk
 from .timeseries import MvStudentTRandomWalk
 
+from .bart import BART
+
 from .bound import Bound
 
+
 __all__ = [
     "Uniform",
     "Flat",
@@ -177,4 +180,5 @@
     "Moyal",
     "Simulator",
     "fast_sample_posterior_predictive",
+    "BART",
 ]