Tech moves fast! Stay ahead of the curve with Techopedia!
Join nearly 200,000 subscribers who receive actionable tech insights from Techopedia.
Deep Q Networks (DQN) are neural networks (and/or related tools) that utilize deep Q learning in order to provide models such as the simulation of intelligent video game play. Rather than being a specific name for a specific neural network build, Deep Q Networks may be composed of convolutional neural networks and other structures that use specific methods to learn about various processes.
The method of deep Q learning typically uses something called general policy iteration, described as the conjunction of policy evaluation and policy iteration, to learn policies from high dimensional sensory input.
For example, a common type of deep Q network covered in tech publications like Medium takes sensory input from Atari 2600 video games to model outcomes. This is done on a very fundamental level by gathering samples, storing them and using them for experience replay in order to update the Q network.
In a general sense, deep Q networks train on inputs that represent active players in areas or other experienced samples and learn to match those data with desired outputs. This is a powerful method in the development of artificial intelligence that can play games like chess at a high level, or carry out other high-level cognitive activities – the Atari or chess video game playing example is also a good example of how AI uses the types of interfaces that were traditionally used by human agents.
In other words, with deep Q learning, the AI player gets to be more like a human player in learning to achieve desired outcomes.