Merge pull request #191 from UCL-CCS/adapt_bugfix

Fixes to subgraph isomorphism utils

symmer/evolution/utils.py

@@ -34,7 +34,8 @@ def get_CNOT_connectivity_graph(evolution_obj:Union[PauliwordOp,QuantumCircuit],
     return G
 
 def _subgraph_isomorphism_distance(G, target, depth=0):
-
+    """
+    """
     if depth == 0:
         if GraphMatcher(target, G).subgraph_is_isomorphic():
             return 0
@@ -56,17 +57,28 @@ def _subgraph_isomorphism_distance(G, target, depth=0):
         return None
 
 def subgraph_isomorphism_distance(G, target, max_depth=3):
+    """
+    """
     depth = 0
     for depth in range(max_depth):
         dist = _subgraph_isomorphism_distance(G, target, depth)
         if dist is not None:
-            return depth * dist
+            return dist
         else:
             depth += 1
     return None
 
 def topology_match_score(ansatz_operator, topology, max_depth=3):
-    entangling_graph = get_CNOT_connectivity_graph(ansatz_operator)
-    subgraph_cost = subgraph_isomorphism_distance(entangling_graph, topology, max_depth=max_depth)
-    n_entangling_gates = np.count_nonzero(ansatz_operator.X_block | ansatz_operator.Z_block)
-    return 1-subgraph_cost/n_entangling_gates
+    """
+    """
+    n_entangling_gates = 2*(np.count_nonzero(ansatz_operator.X_block | ansatz_operator.Z_block)-ansatz_operator.n_terms)
+    if n_entangling_gates == 0:
+        return 1
+    else:
+        entangling_graph = get_CNOT_connectivity_graph(ansatz_operator)
+        subgraph_cost = subgraph_isomorphism_distance(entangling_graph, topology, max_depth=max_depth)
+        if subgraph_cost is None:
+            return 0
+        else:
+            return 1-subgraph_cost/n_entangling_gates
+

symmer/evolution/variational_optimization.py

@@ -27,7 +27,7 @@ class VQE_Driver:
         expectation_eval (str): expectation value method. Its default value is 'symbolic_direct'.
         verbose (bool): If True, prints out useful information during computation. By default it is set to 'True'.
     """
-    expectation_eval = 'symbolic_direct'
+    expectation_eval = 'sparse_array'
     # prints out useful information during computation:
     verbose = True
 
@@ -240,7 +240,7 @@ class ADAPT_VQE(VQE_Driver):
     # method by which to calculate the operator pool derivatives, either
     # commutators: compute the commutator of the observable with each pool element
     # param_shift: use the parameter shift rule, requiring two expectation values per derivative
-    derivative_eval = 'commutators'
+    derivative_eval = 'param_shift'
     # we have alost implemented TETRIS-ADAPT-VQE as per https://doi.org/10.48550/arXiv.2209.10562
     # that aims to reduce circuit-depth in the ADAPT routine by adding multiple excitation terms
     # per cycle that are supported on distinct qubit positions.