def expected_performance(x):
"""Example of how the expected performance can be computed."""
pm_pars = {'alpha': x[0], 'beta': x[1]}
rm_pars = {'theta': x[2]}
T = 100 #number of trials
n_b = 2 #number of bandits
rho = 0.01 #switch probability of the arm-reward contingencies
n_env = 100 #number of environments
n_blocks = 1 #number of experimental blocks
ep = 0 # expected performance
for n in range(n_env):
#generate n_env mutli-armed bandit environmets
env = MultiArmedBandit(T, rho=rho, n_b=n_b)
pm = RescorlaWagner(env, n_b)
for m in range(n_blocks):
#in each environment repeat the experiment n_blocks times
rm = SoftMaxResponses([], pm, d_r)
rm.get_responses(pm_pars, rm_pars)
#For each block compute the expected performance.
ep += env.expected_performance()
return ep / (n_env * n_blocks)
def expected_performance(x):
"""Example of how the expected performance can be computed."""
pm_pars = {'alpha': x[0], 'beta': x[1]}
rm_pars = {'theta': x[2]}
T = 100 #number of trials
n_b = 2 #number of bandits
rho = 0.01 #switch probability of the arm-reward contingencies
n_env = 100 #number of environments
n_blocks = 1 #number of experimental blocks
ep = 0 # expected performance
for n in range(n_env):
#generate n_env mutli-armed bandit environmets
env = MultiArmedBandit(T, rho = rho, n_b = n_b)
pm = RescorlaWagner(env, n_b)
for m in range(n_blocks):
#in each environment repeat the experiment n_blocks times
rm = SoftMaxResponses([], pm, d_r)
rm.get_responses(pm_pars, rm_pars)
#For each block compute the expected performance.
ep += env.expected_performance()
return ep/(n_env*n_blocks)
