diff --git a/docs/src/assets/map_combinator.png b/docs/src/assets/map_combinator.png new file mode 100644 index 00000000..53716d57 Binary files /dev/null and b/docs/src/assets/map_combinator.png differ diff --git a/docs/src/assets/recurse_combinator.png b/docs/src/assets/recurse_combinator.png new file mode 100644 index 00000000..3159ab5a Binary files /dev/null and b/docs/src/assets/recurse_combinator.png differ diff --git a/docs/src/assets/rjmcmc.png b/docs/src/assets/rjmcmc.png new file mode 100644 index 00000000..ce1cdb8d Binary files /dev/null and b/docs/src/assets/rjmcmc.png differ diff --git a/docs/src/assets/scaling_with_sml_graph.png b/docs/src/assets/scaling_with_sml_graph.png new file mode 100644 index 00000000..419a13f3 Binary files /dev/null and b/docs/src/assets/scaling_with_sml_graph.png differ diff --git a/docs/src/assets/static_graph.png b/docs/src/assets/static_graph.png new file mode 100644 index 00000000..8fcdb1ea Binary files /dev/null and b/docs/src/assets/static_graph.png differ diff --git a/docs/src/assets/switch_combinator.png b/docs/src/assets/switch_combinator.png new file mode 100644 index 00000000..98c4e68e Binary files /dev/null and b/docs/src/assets/switch_combinator.png differ diff --git a/docs/src/assets/unfold_combinator.png b/docs/src/assets/unfold_combinator.png new file mode 100644 index 00000000..8fa1b694 Binary files /dev/null and b/docs/src/assets/unfold_combinator.png differ diff --git a/docs/src/ref/inference/mcmc.md b/docs/src/ref/inference/mcmc.md index a36332b2..9e092482 100644 --- a/docs/src/ref/inference/mcmc.md +++ b/docs/src/ref/inference/mcmc.md @@ -345,7 +345,7 @@ end ``` We can then compare the results to the results from the Markov chain that used the selection-based structure-changing kernel: -![rjmcmc plot](../../assets/rjmcmc.png) +![rjmcmc plot](../../../assets/rjmcmc.png) We see that if we initialize the Markov chains from the same state with a single mean (`:z` is `false`) then the selection-based kernel fails to accept any moves to the two-mean structure within 100 iterations, whereas the split-merge kernel transitions back and forth many times, If we repeated the selection-based kernel for enough iterations, it would eventually transition back and forth at the same rate as the split-merge. diff --git a/docs/src/ref/modeling/combinators.md b/docs/src/ref/modeling/combinators.md index 5d7967d7..9b2fda3d 100644 --- a/docs/src/ref/modeling/combinators.md +++ b/docs/src/ref/modeling/combinators.md @@ -14,7 +14,7 @@ Map In the schematic below, the kernel is denoted ``\mathcal{G}_{\mathrm{k}}``. ```@raw html
- schematic of map combinator + schematic of map combinator
``` @@ -61,7 +61,7 @@ In the schematic below, the kernel is denoted ``\mathcal{G}_{\mathrm{k}}``. The initial state is denoted ``y_0``, the number of applications is ``n``, and the remaining arguments to the kernel not including the state, are ``z``. ```@raw html
- schematic of unfold combinator + schematic of unfold combinator
``` @@ -118,7 +118,7 @@ Recurse ```@raw html
- schematic of recurse combinatokr + schematic of recurse combinatokr
``` ## Switch combinator @@ -129,7 +129,7 @@ Switch ```@raw html
- schematic of switch combinator + schematic of switch combinator
``` diff --git a/docs/src/tutorials/scaling_with_sml.md b/docs/src/tutorials/scaling_with_sml.md index 261173d4..3b039c4d 100644 --- a/docs/src/tutorials/scaling_with_sml.md +++ b/docs/src/tutorials/scaling_with_sml.md @@ -244,7 +244,7 @@ Below, we show the static dependency graph that Gen builds for this function. Ar ```@raw html
- static dependency graph + static dependency graph
```