Fix symmetry enforcement with sparse matrices in Laplacian filter (#52)

* laplacian error fix

* Revert "laplacian error fix"

This reverts commit 63c73e0.

* Convert sparse adjacency matrix to dense

Prevents error when enforcing symmetry

Co-authored-by: Kevin Lane <[email protected]>

Author: aaditgupta21

torchts/utils/graph.py

@@ -30,6 +30,9 @@ def reverse_random_walk(adj_mx):
 
 
 def scaled_laplacian(adj_mx, lambda_max=2, undirected=True):
+    if sp.issparse(adj_mx):
+        adj_mx = adj_mx.todense()
+
     if undirected:
         adj_mx = np.maximum.reduce([adj_mx, adj_mx.T])