পদ্ধতির তুলনা করুন

	পলিসি গ্রেডিয়েন্ট পদ্ধতি	গভীর শক্তিশালীকরণ শিক্ষা
ক্ষেত্র	যন্ত্র শিখন	গভীর শিখন
পরিবার	Machine learning	Machine learning
উদ্ভবের বছর	1992	2015
প্রবর্তক	Ronald Williams (REINFORCE); Sutton et al. (policy gradient theorem)	Mnih, V. et al. (DQN)
ধরন	Policy-based reinforcement learning	Sequential decision-making (agent–environment interaction)
মৌলিক উৎস	Williams, R. J. (1992). Simple statistical gradient-following algorithms for connectionist reinforcement learning. Machine Learning, 8(3–4), 229–256.	Mnih, V. et al. (2015). Human-Level Control through Deep Reinforcement Learning. Nature, 518, 529–533.
অপর নাম	REINFORCE, actor-critic, policy optimization, politika gradyanı	Derin Pekiştirmeli Öğrenme (DQN / PPO / A3C), derin pekiştirmeli öğrenme, deep RL, DRL
সম্পর্কিত	4	4
সারসংক্ষেপ	Policy gradient methods are reinforcement-learning algorithms that optimize a parameterized policy directly by gradient ascent on the expected return, rather than learning action-values and acting greedily. Founded on Ronald Williams' 1992 REINFORCE algorithm and the policy gradient theorem of Sutton and colleagues (2000), they naturally handle stochastic and continuous action spaces and underpin modern actor-critic and deep-RL algorithms.	Deep Reinforcement Learning combines neural networks with reinforcement learning so an agent learns by interacting with an environment, popularised by Mnih and colleagues' 2015 Nature work on human-level Atari control. Instead of learning from a fixed labelled dataset, the agent takes actions, observes rewards, and gradually shapes a policy that maximises long-run return.