Merge pull request #613 from frankschae/bb_docs

Update note on sensitivity analysis for hybrid DEs
SciML · Aug 27, 2021 · 4bcfc23 · 4bcfc23
2 parents c4f2ddb + 36fc6f7
commit 4bcfc23
Show file tree

Hide file tree

Showing 2 changed files with 20 additions and 10 deletions.
diff --git a/docs/src/examples/bouncing_ball.md b/docs/src/examples/bouncing_ball.md
@@ -53,8 +53,13 @@ an optimal drag coefficient.
 
 ## Note on Sensitivity Methods
 
-Only some continuous adjoint sensitivities are compatible with callbacks, namely
-`BacksolveAdjoint` and `InterpolatingAdjoint`. All methods based on discrete sensitivity
-analysis via automatic differentiation, like `ReverseDiffAdjoint`, `TrackerAdjoint`, or
-`ForwardDiffAdjoint` are the methods to use (and `ReverseDiffAdjoint` is demonstrated above),
-are compatible with events. This applies to SDEs, DAEs, and DDEs as well.
+The continuous adjoint sensitivities `BacksolveAdjoint`, `InterpolatingAdjoint`,
+and `QuadratureAdjoint` are compatible with events for ODEs. `BacksolveAdjoint` and
+`InterpolatingAdjoint` can also handle events for SDEs. Use `BacksolveAdjoint` if
+the event terminates the time evolution and several states are saved. Currently,
+the continuous adjoint sensitivities do not support multiple events per time point.
+
+All methods based on discrete sensitivity analysis via automatic differentiation,
+like `ReverseDiffAdjoint`,m`TrackerAdjoint`, or `ForwardDiffAdjoint` are the methods
+to use (and `ReverseDiffAdjoint` is demonstrated above), are compatible with events.
+This applies to SDEs, DAEs, and DDEs as well.
diff --git a/docs/src/examples/hybrid_diffeq.md b/docs/src/examples/hybrid_diffeq.md
@@ -71,8 +71,13 @@ Flux.train!(loss_n_ode, ps, data, ADAM(0.05), cb = cba)
 
 ## Note on Sensitivity Methods
 
-Only some continuous adjoint sensitivities are compatible with callbacks, namely
-`BacksolveAdjoint` and `InterpolatingAdjoint`. All methods based on discrete sensitivity
-analysis via automatic differentiation, like `ReverseDiffAdjoint`, `TrackerAdjoint`, or
-`ForwardDiffAdjoint` are the methods to use (and `ReverseDiffAdjoint` is demonstrated above),
-are compatible with events. This applies to SDEs, DAEs, and DDEs as well.
+The continuous adjoint sensitivities `BacksolveAdjoint`, `InterpolatingAdjoint`,
+and `QuadratureAdjoint` are compatible with events for ODEs. `BacksolveAdjoint` and
+`InterpolatingAdjoint` can also handle events for SDEs. Use `BacksolveAdjoint` if
+the event terminates the time evolution and several states are saved. Currently,
+the continuous adjoint sensitivities do not support multiple events per time point.
+
+All methods based on discrete sensitivity analysis via automatic differentiation,
+like `ReverseDiffAdjoint`,m`TrackerAdjoint`, or `ForwardDiffAdjoint` are the methods
+to use (and `ReverseDiffAdjoint` is demonstrated above), are compatible with events.
+This applies to SDEs, DAEs, and DDEs as well.