search

stianteien / TrainsWithBrains

0 stars 0 forks source link

ddqp - monitor loss #10

Open stianteien opened 1 year ago

stianteien commented 1 year ago

Monitor loss for all agents to see if it learns something new.

stianteien commented 1 year ago

Loss not symmetric for the same actions..? image

stianteien commented 1 year ago

Print y_target as well, to see the loss distance from right answer.

stianteien commented 1 year ago

Same loss all the time due to random learning patterns from the memory.