This research focuses on the advanced machine learning study about "Learning to learn", to automatically design contextual-aware neural network for different perceived data with high-efficiency models given specific computer vision tasks.
Public datasets including human actions, emotions and bio-signal neuroimaging will be used for experiments.
Three key problems are to be addressed: How to get machines have the self-learning capability so that they could automatically build deep neural networks for different input with high dimensions? 2) How to automatically build the deep architecture for analyzing the contextual information? 3) How to generate a compact model for decreasing the computational and storage costs?
The research results are expected to provide general solutions and can be applied to any learning tasks with neural network design in portable devices.
The research will be carried out in the Center for Machine Vision and Signal Analysis, University of Oulu.