Fixed a bug on coping ndarray. #1
Conversation
…pied and so modifing newStateValue affected stateValue also. As a result, np.sum(np.abs(newStateValue - stateValue)) was always 0 at the 2nd time of the loop.
|
Thanks very much for pointing this out. I just came from C++. Once I thought numpy has an overload for = operation between ndarray ... It seems I need to check all my code. Hope I don't need too much changes, I remember I also use np.copy in many places. Thanks again! |
|
Thanks god it seems only this example has this bug. If I reinitialize newStateValue in while loop, this bug can also be avoided -- Before using np.copy, I did so in many other examples. Thanks a lot! |
|
Thank you for merging the code. The project is really helpful to understand the book and reinforcement learning. |
|
Hi kentan, would you mind giving me your name and email address? I wanna add you to the contributor list at the header of this file |
|
That would be nice. |
Update grid_world.py to include exact value computations
Introduction: Fix a typo
Update home repo
At current code, just reference was copied.
and so modifying newStateValue affected stateValue also.
As a result, np.sum(np.abs(newStateValue - stateValue)) was always 0 at the 2nd time of the loop.