EasyDeepLearn

TD3 — Twin Delayed DDPG improvements over DDPG.

hard

Answer

  • Fujimoto et al. 2018 fixes: (1) Twin critics: two Q-networks, use min for target → mitigates overestimation.
  • (2) Delayed policy updates: update actor every 2 critic steps → gives critic time to catch up.
  • (3) Target policy smoothing: add clipped noise to target actions → reduces exploitation of Q's peaks.
  • Result: much more stable than DDPG on continuous control.
  • Standard baseline before SAC.
Check yourself — multiple choice
  • Same as DDPG
  • Twin Q (min for target, avoids overestimation) + delayed policy updates + target policy noise smoothing → stable continuous control; SAC/TD3 replaced DDPG
  • Random
  • Not real

TD3: twin Q + delayed policy + target smoothing → stable continuous control.

#policy-methods#actor-critic#deep-rl

Practise Reinforcement Learning

214 interview questions in this topic.

Related questions