Merge branch 'master' into smc_ml

Author: aloctavodia

README.rst

@@ -51,8 +51,8 @@ Learn Bayesian statistics with a book together with PyMC3:
 
 -  `Probabilistic Programming and Bayesian Methods for Hackers <https://github.com/CamDavidsonPilon/Probabilistic-Programming-and-Bayesian-Methods-for-Hackers>`__: Fantastic book with many applied code examples.
 -  `PyMC3 port of the book "Doing Bayesian Data Analysis" by John Kruschke <https://github.com/aloctavodia/Doing_bayesian_data_analysis>`__ as well as the `second edition <https://github.com/JWarmenhoven/DBDA-python>`__: Principled introduction to Bayesian data analysis.
--  `PyMC3 port of the book "Statistical Rethinking A Bayesian Course with Examples in R and Stan" by Richard McElreath <https://github.com/aloctavodia/Statistical-Rethinking-with-Python-and-PyMC3>`__
--  `PyMC3 port of the book "Bayesian Cognitive Modeling" by Michael Lee and EJ Wagenmakers <https://github.com/junpenglao/Bayesian-Cognitive-Modeling-in-Pymc3>`__: Focused on using Bayesian statistics in cognitive modeling.
+-  `PyMC3 port of the book "Statistical Rethinking A Bayesian Course with Examples in R and Stan" by Richard McElreath <https://github.com/pymc-devs/resources/tree/master/Rethinking>`__
+-  `PyMC3 port of the book "Bayesian Cognitive Modeling" by Michael Lee and EJ Wagenmakers <https://github.com/pymc-devs/resources/tree/master/BCM>`__: Focused on using Bayesian statistics in cognitive modeling.
 -  `Bayesian Analysis with Python by Osvaldo Martin <https://www.packtpub.com/big-data-and-business-intelligence/bayesian-analysis-python>`__ (and `errata <https://github.com/aloctavodia/BAP>`__): Great introductory book.
 
 PyMC3 talks

RELEASE-NOTES.md

@@ -6,6 +6,7 @@
 - Renamed `sample_ppc()` and `sample_ppc_w()` to `sample_posterior_predictive()` and `sample_posterior_predictive_w()`, respectively.
 - Refactor SMC and properly compute marginal likelihood (#3124)
 
+
 ## PyMC 3.5 (July 21 2018)
 
 ### New features

build_and_deploy_docs.sh

@@ -1,5 +1,8 @@
 #!/bin/sh
 
+latesttag=$(git describe --tags `git rev-list --tags --max-count=1`)
+echo checking out ${latesttag}
+git checkout ${latesttag}
 pushd docs/source
 make html
 ghp-import -c docs.pymc.io -n -p _build/html/

docs/source/advanced_theano.rst

@@ -60,9 +60,9 @@ variable for our observations::
       # fit the model
       trace = pm.sample()
 
-      # Switch out the observations and use `sample_ppc` to predict
+      # Switch out the observations and use `sample_posterior_predictive` to predict
       x_shared.set_value([-1, 0, 1.])
-      post_pred = pm.sample_ppc(trace, samples=500)
+      post_pred = pm.sample_posterior_predictive(trace, samples=500)
 
 However, due to the way we handle shapes at the moment, it is
 not possible to change the shape of a shared variable if that would

docs/source/examples.rst

@@ -18,6 +18,7 @@ Howto
    notebooks/PyMC3_tips_and_heuristic.ipynb
    notebooks/LKJ.ipynb
    notebooks/live_sample_plots.ipynb
+   notebooks/sampling_compound_step.ipynb
 
 Applied
 =======

docs/source/intro.rst

@@ -8,7 +8,7 @@ Introduction
 Purpose
 =======
 
-PyMC3 is a probabilistic programming module for Python that allows users to fit Bayesian models using a variety of numerical methods, most notably Markov chain Monte Carlo (MCMC) and variational inference (VI). Its flexibility and extensibility make it applicable to a large suite of problems. Along with core model specification and fitting functionality, PyMC3 includes functionality for summarizing output and for model diagnostics.
+PyMC3 is a probabilistic programming package for Python that allows users to fit Bayesian models using a variety of numerical methods, most notably Markov chain Monte Carlo (MCMC) and variational inference (VI). Its flexibility and extensibility make it applicable to a large suite of problems. Along with core model specification and fitting functionality, PyMC3 includes functionality for summarizing output and for model diagnostics.
 
 
 

docs/source/notebooks/Bayes_factor.ipynb

@@ -437,8 +437,8 @@
    ],
    "source": [
     "_, ax = plt.subplots()\n",
-    "ppc_0 = pm.sample_ppc(traces[0], 1000, models[0], size=(len(y), 20))\n",
-    "ppc_1 = pm.sample_ppc(traces[1], 1000, models[1], size=(len(y), 20))\n",
+    "ppc_0 = pm.sample_posterior_predictive(traces[0], 1000, models[0], size=(len(y), 20))\n",
+    "ppc_1 = pm.sample_posterior_predictive(traces[1], 1000, models[1], size=(len(y), 20))\n",
     "for m_0, m_1 in zip(ppc_0['yl'].T, ppc_1['yl'].T):\n",
     "    pm.kdeplot(np.mean(m_0, 0), ax=ax, color='C0', label='model 0')\n",
     "    pm.kdeplot(np.mean(m_1, 0), ax=ax, color='C1', label='model 1')\n",

docs/source/notebooks/Euler-Maruyama and SDEs.ipynb renamed to docs/source/notebooks/Euler-Maruyama_and_SDEs.ipynb

@@ -431,7 +431,7 @@
    ],
    "source": [
     "# generate trace from posterior\n",
-    "ppc_trace = pm.sample_ppc(trace, model=model)\n",
+    "ppc_trace = pm.sample_posterior_predictive(trace, model=model)\n",
     "\n",
     "# plot with data\n",
     "figure(figsize=(10, 3))\n",
@@ -799,7 +799,7 @@
    ],
    "source": [
     "# generate trace from posterior\n",
-    "ppc_trace = pm.sample_ppc(trace, model=model)\n",
+    "ppc_trace = pm.sample_posterior_predictive(trace, model=model)\n",
     "\n",
     "# plot with data\n",
     "figure(figsize=(10, 3))\n",

docs/source/notebooks/GLM-robust-with-outlier-detection.ipynb

@@ -301,7 +301,7 @@
     "\n",
     "### Using `.conditional`\n",
     "\n",
-    "Next, we extend the model by adding the conditional distribution so we can predict at new $x$ locations.  Lets see how the extrapolation looks out to higher $x$.  To do this, we extend our `model` with the `conditional` distribution of the GP.  Then, we can sample from it using the `trace` and the `sample_ppc` function.  This is similar to how Stan uses its `generated quantities {...}` blocks.  We could have included `gp.conditional` in the model *before* we did the NUTS sampling, but it is more efficient to separate these steps."
+    "Next, we extend the model by adding the conditional distribution so we can predict at new $x$ locations.  Lets see how the extrapolation looks out to higher $x$.  To do this, we extend our `model` with the `conditional` distribution of the GP.  Then, we can sample from it using the `trace` and the `sample_posterior_predictive` function.  This is similar to how Stan uses its `generated quantities {...}` blocks.  We could have included `gp.conditional` in the model *before* we did the NUTS sampling, but it is more efficient to separate these steps."
    ]
   },
   {
@@ -334,7 +334,7 @@
     "\n",
     "# Sample from the GP conditional distribution\n",
     "with model:\n",
-    "    pred_samples = pm.sample_ppc(trace, vars=[f_pred], samples=1000)"
+    "    pred_samples = pm.sample_posterior_predictive(trace, vars=[f_pred], samples=1000)"
    ]
   },
   {
@@ -543,7 +543,7 @@
     "    f_pred = gp.conditional(\"f_pred\", X_new)\n",
     "\n",
     "with model:\n",
-    "    pred_samples = pm.sample_ppc(trace, vars=[f_pred], samples=1000)"
+    "    pred_samples = pm.sample_posterior_predictive(trace, vars=[f_pred], samples=1000)"
    ]
   },
   {