Docs/vi posterior notebook

docs/how_to_guide.rst

@@ -61,6 +61,7 @@ Sampling
 - :doc:`how_to_guide/12_mcmc_diagnostics_with_arviz`
 - :doc:`how_to_guide/23_using_pyro_with_sbi`
 - :doc:`how_to_guide/19_posterior_parameters`
+- :doc:`how_to_guide/24_variational_inference`
 
 
 Diagnostics

docs/how_to_guide/06_choosing_inference_method.ipynb

@@ -51,11 +51,11 @@
    "id": "f91955d8",
    "metadata": {},
    "source": [
-    "**`NPE`** is the only method which estimates the posterior directly. This means that it will not require further sampling steps (such as MCMC) _after_ training. This makes NPE the fastest to `.sample()` from the posterior estimate after training. In addition, it can use an embedding network to learn summary statistics (just like NRE, but unlike NLE), see the tutorial [here]().\n",
+    "**`NPE`** is the only method which estimates the posterior directly. This means that it will not require further sampling steps (such as MCMC) _after_ training. This makes NPE the fastest to `.sample()` from the posterior estimate after training. In addition, it can use an embedding network to learn summary statistics (just like NRE, but unlike NLE), see the tutorial [here](../tutorials/00_getting_started.ipynb).\n",
     "\n",
-    "**`NLE`** learns the likelihood. This allows NLE to emulate the simulator after training. To sample from the posterior, NLE is combined with MCMC or variational inference (see [here]()), which makes it slower to `.sample()`. An advantage of NLE is that, if you have multiple iid observations, then it can be much more simulation efficient than NPE (see tutorial [here]()).\n",
+    "**`NLE`** learns the likelihood. This allows NLE to emulate the simulator after training. To sample from the posterior, NLE is combined with MCMC or [variational inference](24_variational_inference.ipynb), which makes it slower to `.sample()`. An advantage of NLE is that, if you have multiple iid observations, then it can be much more simulation efficient than NPE.\n",
     "\n",
-    "**`NRE`** learns the likelihood-to-evidence ratio. It trains only a classifier (unlike NPE and NLE, which train a generative model), which makes it fastest to train. Like NLE, NRE has to be combined MCMC or variational inference to run `.sample()`. Like NPE, NRE can use an embedding network to learn summary statistics."
+    "**`NRE`** learns the likelihood-to-evidence ratio. It trains only a classifier (unlike NPE and NLE, which train a generative model), which makes it fastest to train. Like NLE, NRE has to be combined with MCMC or [variational inference](24_variational_inference.ipynb) to run `.sample()`. Like NPE, NRE can use an embedding network to learn summary statistics."
    ]
   },
   {

docs/how_to_guide/09_sampler_interface.ipynb

@@ -31,7 +31,7 @@
     "For NLE or NRE:\n",
     "- If you have very few parameters (<3), use `RejectionPosterior` or `MCMCPosterior`\n",
     "- If you have a medium number of parameters (3-10), use `MCMCPosterior`\n",
-    "- If you have many parameters (>10) and the `MCMCPosterior` is too slow, use the `VIPosterior`. Optionally combine the `VIPosterior` with an `ImportanceSamplingPosterior` to improve its accuracy.\n",
+    "- If you have many parameters (>10) and the `MCMCPosterior` is too slow, use the `VIPosterior` (see the [VI guide](24_variational_inference.ipynb)). Optionally combine the `VIPosterior` with an `ImportanceSamplingPosterior` to improve its accuracy.\n",
     "\n",
     "## Overview\n",
     "\n",
@@ -243,7 +243,7 @@
  ],
  "metadata": {
   "kernelspec": {
-   "display_name": "Python 3 (ipykernel)",
+   "display_name": "sbi-1",
    "language": "python",
    "name": "python3"
   },
@@ -257,7 +257,7 @@
    "name": "python",
    "nbconvert_exporter": "python",
    "pygments_lexer": "ipython3",
-   "version": "3.12.4"
+   "version": "3.12.12"
   },
   "toc": {
    "base_numbering": 1,