Thompson sampling on a Bernoulli probabilistic power socket.

BernoulliThompsonSocket.py

class BernoulliThompsonSocket( PowerSocket ):
    def __init__( self, q ):             
                
        self.α = 1  # the number of times this socket returned a charge        
        self.β = 1  # the number of times no charge was returned
        
        # pass the true reward value to the base PowerSocket             
        super().__init__(q)          
    
    def charge(self):        
        """ return some charge with the socket's predefined probability """
        return np.random.random() < self.q
                    
    def update(self,R):
        """ increase the number of times this socket has been used and 
            update the counts of the number of times the socket has and 
            has not returned a charge (alpha and beta)"""
        self.n += 1    
        self.α += R
        self.β += (1-R)
        
    def sample(self):
        """ return a value sampled from the beta distribution """
        return np.random.beta(self.α,self.β)