Fixed issue with storage of fevals in curve data.

mlrose_hiive/algorithms/ga.py

@@ -222,7 +222,7 @@ def genetic_alg(problem, pop_size=200, pop_breed_percent=0.75, elite_dreg_ratio=
             attempts += 1
 
         if curve:
-            fitness_curve.append(problem.get_adjusted_fitness())
+            fitness_curve.append((problem.get_adjusted_fitness(), problem.fitness_evaluations))
 
         # invoke callback
         if state_fitness_callback is not None:

mlrose_hiive/algorithms/gd.py

@@ -109,7 +109,7 @@ def gradient_descent(problem, max_attempts=10, max_iters=np.inf,
             attempts += 1
 
         if curve:
-            fitness_curve.append(problem.get_adjusted_fitness())
+            fitness_curve.append((problem.get_adjusted_fitness(), problem.fitness_evaluations))
 
         # invoke callback
         if state_fitness_callback is not None:

mlrose_hiive/algorithms/hc.py

@@ -103,7 +103,7 @@ def hill_climb(problem, max_iters=np.inf, restarts=0, init_state=None,
             next_fitness = problem.eval_fitness(next_state)
 
             if curve:
-                fitness_curve.append(problem.get_adjusted_fitness())
+                fitness_curve.append((problem.get_adjusted_fitness(), problem.fitness_evaluations))
 
             # invoke callback
             if state_fitness_callback is not None:

mlrose_hiive/algorithms/mimic.py

@@ -136,7 +136,7 @@ def mimic(problem, pop_size=200, keep_pct=0.2, max_attempts=10,
             attempts += 1
 
         if curve:
-            fitness_curve.append(problem.get_adjusted_fitness())
+            fitness_curve.append((problem.get_adjusted_fitness(), problem.fitness_evaluations))
 
         # invoke callback
         if state_fitness_callback is not None:

mlrose_hiive/algorithms/rhc.py

@@ -89,12 +89,15 @@ def random_hill_climb(problem, max_attempts=10, max_iters=np.inf, restarts=0,
     all_curves = []
 
     continue_iterating = True
+    # problem.reset()
     for current_restart in range(restarts + 1):
         # Initialize optimization problem and attempts counter
+        fevals = problem.fitness_evaluations
         if init_state is None:
             problem.reset()
         else:
             problem.set_state(init_state)
+        problem.fitness_evaluations = fevals
 
         fitness_curve = []
         callback_extra_data = None
@@ -128,8 +131,13 @@ def random_hill_climb(problem, max_attempts=10, max_iters=np.inf, restarts=0,
 
             if curve:
                 adjusted_fitness = problem.get_adjusted_fitness()
-                fitness_curve.append(adjusted_fitness)
-                all_curves.append({'current_restart': current_restart, 'Fitness': problem.get_adjusted_fitness()})
+                curve_value = (adjusted_fitness, problem.fitness_evaluations)
+                fitness_curve.append(curve_value)
+                all_curves.append(curve_value)
+                """
+                all_curves.append({'current_restart': current_restart, 'Fitness': problem.get_adjusted_fitness(),
+                                   'FEval': problem.fitness_evaluations})
+                """
 
             # invoke callback
             if state_fitness_callback is not None:
@@ -159,5 +167,4 @@ def random_hill_climb(problem, max_attempts=10, max_iters=np.inf, restarts=0,
             break
     best_fitness *= problem.get_maximize()
 
-
     return best_state, best_fitness, np.asarray(best_fitness_curve) if curve else None

mlrose_hiive/algorithms/sa.py

@@ -128,7 +128,7 @@ def simulated_annealing(problem, schedule=GeomDecay(), max_attempts=10,
                 attempts += 1
 
         if curve:
-            fitness_curve.append(problem.get_adjusted_fitness())
+            fitness_curve.append((problem.get_adjusted_fitness(), problem.fitness_evaluations))
 
         # invoke callback
         if state_fitness_callback is not None:

mlrose_hiive/runners/_runner_base.py

@@ -266,10 +266,12 @@ def _start_run_timing(self):
 
     @staticmethod
     def _create_curve_stat(iteration, curve_value, curve_data, t=None):
+        curve_fitness_value, curve_feval_value = curve_value
         curve_stat = {
             'Iteration': iteration,
             'Time': t,
-            'Fitness': curve_value
+            'Fitness': curve_fitness_value,
+            'FEval': curve_feval_value
         }
 
         curve_stat.update(curve_data)
@@ -310,8 +312,6 @@ def _save_state(self, iteration, state, fitness, user_data,
         current_iteration_stats.update({str(gd(k)): self._sanitize_value(v)
                                         for k, v in {k: v for (k, v) in user_data}.items()})
 
-        if fitness_evaluations is not None:
-            current_iteration_stats['FEvals'] = fitness_evaluations
 
         # check for additional info
         gi = lambda k, v: {} if not hasattr(v, 'get_info__') else v.get_info__(t)
@@ -328,6 +328,7 @@ def _save_state(self, iteration, state, fitness, user_data,
             run_stat = {
                 'Iteration': i,
                 'Fitness': fitness,
+                'FEvals': fitness_evaluations,
                 'Time': t,
                 'State': self._sanitize_value(state)
             }
@@ -338,7 +339,7 @@ def _save_state(self, iteration, state, fitness, user_data,
         if self.generate_curves and iteration == 0:
             # capture first fitness value for iteration 0 if not already captured.
             if curve is None or len(curve) == 0:
-                curve = [fitness]
+                curve = [(fitness, fitness_evaluations)]
                 self._first_curve_synthesized = True
 
         if self.generate_curves and curve is not None:  # and (done or iteration == max(self.iteration_list)):
@@ -364,6 +365,7 @@ def _save_state(self, iteration, state, fitness, user_data,
 
             if self._copy_zero_curve_fitness_from_first and len(self._fitness_curves) > 1:
                 self._fitness_curves[0]['Fitness'] = self._fitness_curves[1]['Fitness']
+                # self._fitness_curves[0]['FEval'] = 0
                 self._copy_zero_curve_fitness_from_first = False
             self._create_and_save_run_data_frames()
 

problem_examples.ipynb

@@ -654,7 +654,7 @@
   {
    "cell_type": "markdown",
    "source": [
-    "The best run using these criteria is as follows:"
+    "The best runs using these criteria is as follows:"
    ],
    "metadata": {
     "collapsed": false
@@ -687,7 +687,8 @@
   {
    "cell_type": "markdown",
    "source": [
-    "Which has the following identifying state information:"
+    "We will arbitrarily pick the first row for this example,\n",
+    "which has the following identifying state information:"
    ],
    "metadata": {
     "collapsed": false