Skip to main content
. Author manuscript; available in PMC: 2020 May 29.
Published in final edited form as: Phys Med Biol. 2019 May 29;64(11):115013. doi: 10.1088/1361-6560/ab18bf

Table 1.

Hyperparameters to train the WTPN.

Hyperparameter Value Description
σ 5×10−4 Stopping criteria in Algorithm 1
β 5 Penalty parameter in Algorithm 1
n 4 Number of weights (OARs) to be tuned
γ 0.5 Discount factor
0.99 ~ 0.1 Probability of -greedy approach
Npatient 5 Number of training patient cases
Nepoch 100 Number of training epoch
Ntrain 25 Number of training steps in each epoch
Nupdate 10 Number of steps to update W^ =W
δ 1×10−4 Learning rate (step size of gradient descent for W)