How Can Deep Reinforcement Learning Be Applied to Intelligent Conversations?

Deep reinforcement learning is one of the research hotspots in the field of artificial intelligence, and its application research progress and results in the field of intelligent dialogue have also attracted a lot of attention.

Deep reinforcement learning

Although some people have tried to combine deep learning and reinforcement learning a long time ago, the beginning of the real success is the article “Playing Atari with Deep Reinforcement Learning” published by DeepMind on NIPS 2013. In 2015, DeepMind’s Volodymyr Mnih and other researchers published a paper “Human-level control through deep learning” in the journal Nature. The paper proposed a model Deep Q-Network (DQN) that combines deep learning technology and reinforcement learning ideas. , Demonstrated performance beyond human level on the Atari game platform. Since then, deep reinforcement learning combining DL and RL has quickly become the focus of the artificial intelligence community.

Generally speaking, reinforcement learning problems in the real world include huge state spaces and action spaces. Traditional reinforcement learning methods are limited by the curse of dimensionality. With the help of neural networks in deep learning, the main body of reinforcement learning can directly…