default line search to False (can be slow)

jameschapman19 · Feb 16, 2023 · 906a2f1 · 906a2f1
1 parent 1c00ed3
commit 906a2f1
Show file tree

Hide file tree

Showing 3 changed files with 9 additions and 5 deletions.
diff --git a/cca_zoo/models/_stochastic/_eigengame.py b/cca_zoo/models/_stochastic/_eigengame.py
@@ -77,7 +77,7 @@ def __init__(
         c=0,
         nesterov=True,
         rho=0.1,
-        line_search=True,
+        line_search=False,
     ):
         super().__init__(
             latent_dims=latent_dims,
@@ -140,7 +140,10 @@ def _Aw(self, view, projections):
         return view.T @ projections / view.shape[0]
 
     def _Bw(self, view, projection, weight, c):
-        return (c * weight) + (1 - c) * (view.T @ projection) / projection.shape[0]
+        if c==1:
+            return (c * weight)
+        else:
+            return (c * weight) + (1 - c) * (view.T @ projection) / projection.shape[0]
 
     def _get_terms(self, i, view, projections, v):
         projections.mask[i] = True
@@ -247,7 +250,7 @@ def __init__(
         epochs=1,
         learning_rate=1,
         nesterov=True,
-        line_search=True,
+        line_search=False,
     ):
         super().__init__(
             latent_dims=latent_dims,
@@ -341,7 +344,7 @@ def __init__(
         epochs=1,
         learning_rate=1,
         nesterov=True,
-        line_search=True,
+        line_search=False,
     ):
         super().__init__(
             latent_dims=latent_dims,

diff --git a/cca_zoo/models/_stochastic/_ghagep.py b/cca_zoo/models/_stochastic/_ghagep.py
@@ -97,6 +97,7 @@ def __init__(
             learning_rate=learning_rate,
             nesterov=nesterov,
             rho=rho,
+            line_search=False,
             c=c
         )
 

diff --git a/examples/plot_dcca.py b/examples/plot_dcca.py
@@ -49,7 +49,7 @@
 # %%
 # Deep CCA EigenGame
 # ----------------------------
-dcca_eg = DCCA_EigenGame(latent_dims=LATENT_DIMS, encoders=[encoder_1, encoder_2])
+dcca_eg = DCCA_EigenGame(latent_dims=LATENT_DIMS, encoders=[encoder_1, encoder_2],lr=1e-5)
 trainer = pl.Trainer(
     max_epochs=EPOCHS,
     enable_checkpointing=False,