Using Image Transformations to Learn Network Structure
This work addresses network design and planning problems, such as shipping logistics, by providing a method to learn and recommend connectivity, but it appears incremental as it adapts existing image compression and reinforcement learning techniques to a specific domain.
The paper tackles the problem of learning network structure from sequences of images by treating networks as images and using image compression to extract geographic signatures, which are then used to recommend future network connectivity; the result is a Bayesian reinforcement algorithm that incorporates these signatures as priors and user decisions to improve probabilistic decision-making.
Many learning tasks require observing a sequence of images and making a decision. In a transportation problem of designing and planning for shipping boxes between nodes, we show how to treat the network of nodes and the flows between them as images. These images have useful structural information that can be statistically summarized. Using image compression techniques, we reduce an image down to a set of numbers that contain interpretable geographic information that we call geographic signatures. Using geographic signatures, we learn network structure that can be utilized to recommend future network connectivity. We develop a Bayesian reinforcement algorithm that takes advantage of statistically summarized network information as priors and user-decisions to reinforce an agent's probabilistic decision. Additionally, we show how reinforcement learning can be used with compression directly without interpretation in simple tasks.