evaluator module

Collects measurements of an agent’s performance.

Authors:

Banafsheh Rafiee, Niko Yasui

class evaluator.Evaluator(gvf_name, num_features, alpha_rupee, beta0_rupee, use_MSRE)

Collects measurements of an agent’s performance

Parameters:

• gvf_name (str) – Name of the GVF used for plotting.
• num_features (int) – Total number of features in the representation.
• alpha_rupee (float) – Primary learning rate for RUPEE.
• beta0_rupee (float) – Initial averaging weight for RUPEE.
• use_MSRE (bool) – Calculate Mean Squared Return Error.

compute_IS_ess(rho, time_step, *args, **kwargs)

compute_MSRE(theta, time_step, *args, **kwargs)

compute_avg_td_error(delta, time_step, *args, **kwargs)

compute_rupee(tderr_elig, phi, *args, **kwargs)

update(*args, **kwargs)