concept relational DQN in category deep learning

appears as: relational DQN, relational DQN
Deep Reinforcement Learning in Action

This is an excerpt from Manning's book Deep Reinforcement Learning in Action.

Overall we will repurpose the relational module we created earlier for the MNIST example as a relational DQN, so we really only need to change the output to a normal activation function rather than the log_softmax we used for classification. But first, let’s get back to implementing multi-head attention. As operating on higher-order tensors gets more complicated, we will get help from a package called Einops that extends the capabilities of PyTorch’s built-in Einsum function. You can install it using pip: