Research in combining reinforcement learning with deep learning can lead to many applications like that in human vision. In this connection the two variants of deep learning; ConvNet and RNN can be combined to develop a model that uses reinforcement learning. It seems that not much work has been done in this direction. Can somebody suggest on effective endeavors in this direction?
You can use a hybrid network containing an RNN and ConvNet to make a nueral net that classifies images of an object which may have varying resolutions.
Actually there are some notable research is going on in this field at Alphabets Deepmind and OpenAI. Also, a new subfield have arrived by the name Deep Reinforcement Learning. Maybe, sir, you will be interested in Deep Q-learning as well.
B.K. Tripathy Dear Professor, indeed, it is a nice research direction. There are some good research works on RNN and RL combination. Some of them used publicly real-world dataset, such as in:
Supervised Reinforcement Learning with Recurrent Neural Network for Dynamic Treatment Recommendation
Link: Preprint Supervised Reinforcement Learning with Recurrent Neural Netw...
- Badges
- Science topic