Merge pull request #28 from deckar01/move-energy-delta

Allow optimizing per-move energy changes

README.md

@@ -135,6 +135,13 @@ them into the constructor like so
 
 The last line (calling init on the super class) is critical. 
 
+## Optimizations
+
+For some problems the `energy` function is prohibitively expensive to calculate
+after every move. It is often possible to compute the change in energy that a
+move causes much more efficiently. A delta value can be returned from `move` to
+update the energy value without calling `energy` multiple times.
+
 ## Implementation Details
 
 The simulated annealing algorithm requires that we track state (current, previous, best) and thus means we need to copy the `self.state` frequently.

simanneal/anneal.py

@@ -193,9 +193,12 @@ def anneal(self):
         while step < self.steps and not self.user_exit:
             step += 1
             T = self.Tmax * math.exp(Tfactor * step / self.steps)
-            self.move()
-            E = self.energy()
-            dE = E - prevEnergy
+            dE = self.move()
+            if dE is None:
+                E = self.energy()
+                dE = E - prevEnergy
+            else:
+                E += dE
             trials += 1
             if dE > 0.0 and math.exp(-dE / T) < random.random():
                 # Restore previous state