From 45f42a28e050a4699ef5e55ee79e78b65b7e196d Mon Sep 17 00:00:00 2001 From: Feda Curic Date: Wed, 27 Dec 2023 09:09:25 +0100 Subject: [PATCH] Update adloc docs State the adloc will be run for every iteration --- .../howto/adaptive_localization.ipynb | 15 ++++++++------- 1 file changed, 8 insertions(+), 7 deletions(-) diff --git a/docs/getting_started/howto/adaptive_localization.ipynb b/docs/getting_started/howto/adaptive_localization.ipynb index 6b65675d180..910b798f4fd 100644 --- a/docs/getting_started/howto/adaptive_localization.ipynb +++ b/docs/getting_started/howto/adaptive_localization.ipynb @@ -169,19 +169,18 @@ "## Adaptive-Localization - Pearson Correlation as measure of closeness\n", "\n", "As mentioned, the difference between distance-based localization and adaptive-localization is the measure of closeness.\n", - "Otherwise, the approach is similar to distance-based localization.\n", + "Otherwise, the approaches are similar.\n", "\n", "- **Adaptive Localization step-by-step:**\n", " - Pick a subset of adjacent parameters\n", " - Pick responses that are significantly correlated to the parameters\n", " - Do the update using only observations that correspond to these responses\n", "\n", - "The primary advantage of adaptive localization is its user-friendliness.\n", - "Users only need to specify the `correlation threshold`.\n", - "Often, the default `correlation threshold` calculated by `ERT` is reasonable.\n", + "The primary advantage of adaptive localization is its user-friendliness as users only need to (optionally) specify the `correlation threshold`.\n", "\n", "To activate adaptive localization in `ERT`, simply check the `Adaptive localization correlation threshold` checkbox.\n", - "The correlation threshold can then be set, though a sensible default will be determined based on ensemble size.\n", + "This activates adaptive localization for the entire experiment, meaning it will run for every iteration of, say, `ES-MDA`.\n", + "While the correlation threshold can be manually set, `ERT` determines a sensible default based on the size of the ensemble.\n", "At a correlation threshold of 0.0 the update will be identical to no localization, while at a correlation threshold of 1.0 the posterior will be identical to the prior, and there will be no update.\n", "\n", "![](ert_screenshot_adaptive_loc.png)" @@ -196,8 +195,10 @@ "\n", "Be aware that adaptive localization comes with significant computational costs, which depend on two factors:\n", "\n", - " - The number of parameters: More parameters lead to longer run times.\n", - " - The correlation threshold: A higher threshold reduces computational time.\n" + "- The number of parameters: More parameters lead to longer run times.\n", + "- The correlation threshold: A higher threshold reduces computational time. \\\n", + " Increasing the threshold makes it less likely that a parameter and a response are considered significantly correlated. \\\n", + " This makes it less likely that a parameter needs to be updated at all which reduces the computational cost.\n" ] }, {