Multi-agent deep reinforcement learning

In this project we will develop deep artificial neural networks architectures for computer vision with applications in digital healthcare and manufacturing.

Main challenge of:

1 imitation learning is: the need of huge amounts of data labelled from experts.
2 reinforcement learning is: to model a reward function (very complex for very complex models, such as real-world scenarios). Doesn't need expert labelled data.

The project is about learning from demonstrations via hybrid approaches (imitation learning, reinforcement learning and transfer learning). This project alligns with EPSRC research areas exploring Artificial Intelligence Technologies