Update to scikit learn 1.0.1 and change the TSNE squared distances de…

…fault.
matthew-gaddes · Sep 20, 2022 · 3e4ac7b · 3e4ac7b
1 parent f319896
commit 3e4ac7b
Show file tree

Hide file tree

Showing 10 changed files with 2 additions and 2 deletions.
diff --git a/ICASAR.yml b/ICASAR.yml
@@ -8,7 +8,7 @@ dependencies:
   - numpy
   - spyder=4
   - matplotlib=3.0.1
-  - scikit-learn=0.20.0
+  - scikit-learn=1.0.1
   - hdbscan=0.8.18
   - simplekml
   - h5py
diff --git a/example_spatial_01_outputs/01_pca_variance_line.png b/example_spatial_01_outputs/01_pca_variance_line.png
diff --git a/example_spatial_01_outputs/02_PCA_sources_correlations.png b/example_spatial_01_outputs/02_PCA_sources_correlations.png
diff --git a/example_spatial_01_outputs/02_PCA_sources_time.png b/example_spatial_01_outputs/02_PCA_sources_time.png
diff --git a/example_spatial_01_outputs/03_ICA_sources_correlations.png b/example_spatial_01_outputs/03_ICA_sources_correlations.png
diff --git a/example_spatial_01_outputs/03_ICA_sources_time.png b/example_spatial_01_outputs/03_ICA_sources_time.png
diff --git a/example_spatial_01_outputs/04_clustering_and_manifold_results.png b/example_spatial_01_outputs/04_clustering_and_manifold_results.png
diff --git a/example_spatial_01_outputs/ICASAR_results.pkl b/example_spatial_01_outputs/ICASAR_results.pkl
diff --git a/example_spatial_01_outputs/ICs.kmz b/example_spatial_01_outputs/ICs.kmz
diff --git a/icasar/icasar_funcs.py b/icasar/icasar_funcs.py
@@ -1053,7 +1053,7 @@ def bootstrapped_sources_to_centrotypes(sources_r2, hdbscan_param, tsne_param):
     # 3:  2d manifold with all the recovered sources
     print('Starting to calculate the 2D manifold representation....', end = "")
     manifold_tsne = TSNE(n_components = 2, metric = 'precomputed', perplexity = perplexity, early_exaggeration = early_exaggeration,
-                         init = 'random', learning_rate = 200.0, square_distances='legacy')                                                                               # default will change to pca in 1.2.  May be worth experimenting with.  
+                         init = 'random', learning_rate = 200.0, square_distances=True)                                                                               # default will change to pca in 1.2.  May be worth experimenting with.  
     xy_tsne = manifold_tsne.fit(D).embedding_
     print('Done!' )