Major update to use TF v2.17. Concrete dropout layers no longer work due to TF/Keras changes, will be fixed later.

Author: mdhimes

LICENSE

@@ -4,9 +4,9 @@ neural network models.
 
 This project was completed with the support of the NASA Fellowship Activity, 
 grant 80NSSC20K0682, held by Principal Investigator Joseph Harrington in 
-support of Michael Himes, the principal developer of HOMER.  Adam D. Cobb,
+support of Michael Himes, the principal developer of MARGE.  Adam D. Cobb,
 David C. Wright, and Zacchaeus Scheffer  contributed to some aspects of the 
-code, as described in the MARGE documentation.
+published version of the code, as described in the MARGE documentation.
 
 -------------------------------------------------------------------------------
 

MARGE.py

@@ -1,21 +1,14 @@
 name: marge
 channels:
-  - defaults
   - conda-forge
 dependencies:
-  - mpi4py=3.0.3
-  - mpich=3.3.2
-  - pip=19.0.3
-  - cython=0.29.5
-  - h5py=2.9.0
-  - hdf5=1.10.4
-  - matplotlib=3.0.2
-  - numpy=1.16.2
-  - python=3.7.2
-  - scikit-learn=0.20.2
-  - scipy=1.2.1
-  - yaml=0.1.7
-  - tensorflow-gpu==1.13.1
-  - keras==2.2.4
-  - pip:
-    - -rrequirements.txt --install-option="--no-deps"
+  - cudnn=9.3.0.75
+  - dask=2024.12.0
+  - h5py=3.12.1
+  - keras=3.7.0
+  - matplotlib=3.10.0
+  - numpy=1.26.4
+  - optuna=4.1.0
+  - scikit-learn=1.6.0
+  - scipy=1.14.1
+  - tensorflow-gpu=2.17.0

doc/MARGE_user_manual/MARGE_user_manual.tex

@@ -1,8 +1,8 @@
 Introduction
 ============
 
-This directory holds an example of how to run MARGE, using BART for data 
-generation. It matches the use case demonstrated in Himes et al. (2020), except
+This directory holds an example of how to run MARGE using BART for data 
+generation. It matches the use case demonstrated in Himes et al. (2022), except
 with reduced accuracy to reduce the runtime.
 
 ********************************** IMPORTANT **********************************