Problem of the final lab

Course Q&A Machine Learning Specialization Unsupervised Learning, Recommenders, Reinforcement
1

image
image808×116 10.9 KB

why is this need，i mean why should the last cost function be different
please help me

2

Hi @wgl

The cost function is the Bellman Equation and it’s being used because we are estimating the optimal action-value function.

This is explained in section 6 - Deep Q-Learning of the lab.

DeepQLearning
DeepQLearning1007×340 39.8 KB

QNetwork
QNetwork1142×656 55.3 KB

Sam

3

thanks，but i am still comfuse about why if episode terminate at j+1 and yj = Rj but not yj=
Rj+rmaxQ

4

I think all it’s saying is that if you’re at the last state, there’s no “j+1” possible, so you ignore it.