swarmrl.value_functions.generalized_advantage_estimate Module API Reference¶
Module for the expected returns value function.
GAE
¶
Class for the expected returns.
Source code in swarmrl/value_functions/generalized_advantage_estimate.py
14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 |
|
__call__(rewards, values)
¶
Call function for the advantage. Parameters
rewards : np.ndarray (n_time_steps, n_particles) A numpy array of rewards to use in the calculation. values : np.ndarray (n_time_steps, n_particles) The prediction of the critic for the episode. Returns
advantages : np.ndarray (n_time_steps, n_particles) Expected returns for the rewards.
Source code in swarmrl/value_functions/generalized_advantage_estimate.py
41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 |
|
__init__(gamma=0.99, lambda_=0.95)
¶
Constructor for the generalized advantage estimate class
Parameters¶
gamma : float A decay factor for the values of the task each time step. lambda_ : float A decay factor that describes the amount of bias included in the advantage calculation.
Notes¶
See https://arxiv.org/pdf/1506.02438.pdf for more information.
Source code in swarmrl/value_functions/generalized_advantage_estimate.py
19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 |
|