def work()

in isoexp/linear/linearbandit.py [0:0]

• 13 lines of code
• 4 McCabe index (conditional complexity)

        def work(alg_name, alg):

            regret = np.zeros((nb_simu, T))
            norms = np.zeros((nb_simu, T))

            for k in trange(nb_simu, desc='Simulating {}'.format(alg_name)):

                alg.reset()

                for t in trange(T, desc='Current episode :', leave=True):
                    a_t = alg.get_action()
                    # print(a_t)
                    r_t = model.reward(a_t)
                    alg.update(a_t, r_t)

                    regret[k, t] = link(model.best_arm_reward()) - link(np.dot(model.theta, model.features[a_t]))
                    if hasattr(alg, 'theta_hat'):
                        norms[k, t] = np.linalg.norm(alg.theta_hat - model.theta, 2)

            return alg_name, MABResults(regret=regret, norm_error=norms)