Added multiple trial functionality, other minor changes

slots/slots.py

@@ -39,7 +39,7 @@ def __init__(self, num_bandits=3, probs=None, payouts=None, live=True,
         stop_criterion : dict
             Stopping criterion (str) and threshold value (float).
         '''
-
+        self.num_bandits = num_bandits
         self.choices = []
 
         if not probs:
@@ -327,20 +327,47 @@ def best(self):
         else:
             return np.argmax(self.wins/(self.pulls+0.1))
 
-    def est_payouts(self):
+    def current(self):
         '''
-        Calculate current estimate of average payout for each bandit.
+        Return last choice of bandit.
 
         Returns
         -------
-        array of floats or None
+        int
+            Index of bandit
         '''
 
         if len(self.choices) < 1:
             print('slots: No trials run so far.')
             return None
         else:
-            return self.wins/(self.pulls+0.1)
+            return self.choices[-1]
+
+    def est_payouts(self, bandit=None):
+        '''
+        Calculate current estimate of average payout for each bandit.
+
+        Parameters
+        ----------
+        bandit : None
+            If a bandit is selected, return the payout for that bandit, otherwise return all payouts. 
+
+        Returns
+        -------
+        array of floats or None
+        '''
+        if not bandit:
+            if len(self.choices) < 1:
+                print('slots: No trials run so far.')
+                return None
+            else:
+                return self.wins/(self.pulls+0.1)
+        else:
+            if len(self.choices) < 1:
+                print('slots: No trials run so far.')
+                return None
+            else:
+                return (self.wins/(self.pulls+0.1))[bandit]
 
     def regret(self):
         '''
@@ -430,6 +457,95 @@ def online_trial(self, bandit=None, payout=None, strategy='eps_greedy',
             return {'new_trial': True,
                     'choice': self.run_strategy(strategy, parameters),
                     'best': self.best()}
+
+
+    # The next two are my implementations of inserting a whole array into the model so you're not manually iterating it
+    def multiple_trials(self, bandits=None, payouts=None, strategy='eps_greedy',
+                     parameters=None):
+        '''
+        Feeds two arrays in and based on those results returns the next trial. 
+        This really isn't optimized, there's a much better way of doing this if we don't 
+        care about maintaining the 
+
+        Parameters
+        ----------
+        bandit : array of ints
+            Bandit index
+        payout : array of floats
+            Payout value
+        strategy : string
+            Name of update strategy
+        parameters : dict
+            Parameters for update strategy function
+
+        Returns
+        -------
+        dict
+            Format: {'new_trial': boolean, 'choice': int, 'best': int}
+        '''
+        if len(payouts) is not len(bandits):
+            raise Exception('slots.online_trials: number of bandits is different from number of payouts')
+        else:
+            for x in range(0,len(payouts)):
+                if bandits[x] is not None and payouts[x] is not None:
+                    self.update(bandit=bandits[x], payout=payouts[x])
+                else:
+                    raise Exception('slots.online_trial: bandit and/or payout value'
+                            ' missing.')
+
+        if self.crit_met():
+            return {'new_trial': False, 'choice': self.best(),
+                    'best': self.best()}
+        else:
+            return {'new_trial': True,
+                    'choice': self.run_strategy(strategy, parameters),
+                    'best': self.best()}
+
+
+
+    def multiple_trials_bulk(self, bandits=None, payouts=None, strategy='eps_greedy',
+                     parameters=None):
+        '''
+        Feeds two arrays in and based on those results returns the next trial. 
+        This really isn't optimized, there's a much better way of doing this if we don't 
+        care about maintaining the 
+
+        Parameters
+        ----------
+        bandit : array of ints
+            Bandit index
+        payout : array of floats
+            Payout value
+        strategy : string
+            Name of update strategy
+        parameters : dict
+            Parameters for update strategy function
+
+        Returns
+        -------
+        dict
+            Format: {'new_trial': boolean, 'choice': int, 'best': int}
+        '''
+
+        #do I need an error here in case someone puts a string or some shit?
+        banditos = np.array(bandits)
+        if len(payouts) is not len(bandits):
+            raise Exception('slots.online_trials: number of bandits is different from number of payouts')
+        else:
+            self.choices.append(bandits)
+            for y in list(set(bandits)):
+                indices = np.where(banditos == y)[0]
+                payola = [payouts[i] for i in indices]
+                self.pulls[y] += len(payola)
+                self.wins[y] += sum(payola)
+                self.bandits.payouts[y] += sum(payola)
+        if self.crit_met():
+            return {'new_trial': False, 'choice': self.best(),
+                    'best': self.best()}
+        else:
+            return {'new_trial': True,
+                    'choice': self.run_strategy(strategy, parameters),
+                    'best': self.best()}
 
     def update(self, bandit, payout):
         '''
@@ -451,6 +567,11 @@ def update(self, bandit, payout):
         self.wins[bandit] += payout
         self.bandits.payouts[bandit] += payout
 
+    def info(self):
+        '''
+        Default: display number of bandits, wins, and estimated probabilities
+        '''
+        return('number of bandits:',self.num_bandits, 'number of wins', self.wins, 'estimated payouts', self.est_payouts())
 
 class Bandits():
     '''
@@ -511,5 +632,3 @@ def pull(self, i):
             else:
                 return 0.0
 
-    def info(self):
-        pass