agent.py 文件源码

def run_single_softmax_experiment(beta, alpha):
    """Run experiment with agent using softmax update rule."""
    print('Running a contextual bandit experiment')
    cb = ContextualBandit()
    ca = ContextualAgent(cb, beta=beta, alpha=alpha)
    trials = 360

    for _ in range(trials):
        ca.run()
    df = DataFrame(ca.log, columns=('context', 'action', 'reward', 'Q(c,23)',
                                    'Q(c,14)', 'Q(c,8)', 'Q(c,3)'))
    # fn = 'softmax_experiment.csv'
    # df.to_csv(fn, index=False)
    # print('Sequence written in', fn)
    # globals().update(locals())  #

    return df