summaryrefslogtreecommitdiff
diff options
context:
space:
mode:
authorLoïc Guégan <loic.guegan@mailbox.org>2025-09-23 13:57:45 +0200
committerLoïc Guégan <loic.guegan@mailbox.org>2025-09-23 13:57:45 +0200
commit436db170f88ae1f14c1fdf36518ead9216abe80f (patch)
treeff377ae41ed577f3191104bd29167331cf54cd15
parentebb8ea1e9647728827cba1fd38915f06a86bcdc2 (diff)
Update README.md
-rw-r--r--README.md18
1 files changed, 18 insertions, 0 deletions
diff --git a/README.md b/README.md
index 7e761a1..830765c 100644
--- a/README.md
+++ b/README.md
@@ -9,3 +9,21 @@
- Contains the simulator source code and script to run the experiments from the paper.
- `slides_recap/`
- Just backup slides. Nothing interresting.
+
+## Experiments
+
+### In-situ
+
+Simulations are carried using previous CCGrid2022 date in the file `analysis/in-situ.R`.
+The idea is to use a single policy each day (round-robin) until model performance reaches 89%.
+For that, a day correspond to a single simulation from CCGrid2022 with a particular policy.
+Since we have 200 runs per policy from CCGrid2022, a single run result can be cherry-pick to simulate one day.
+
+### Offline
+
+First, models are trained and predictions are generated.
+This is done in `analysis/offline.R`.
+Predictions are stored in `analysis/inputs/input_*.csv`.
+Then simulation are executed with those policies predictions as inputs.
+This is done in `simulations/results/paper.sh`.
+