Recently, there is a rising interest in applying neural information processing to solve real-life problems and to model how the brain works. In this term, we will focus on highway traffic data. It shares a lot of characteristics with granular flows in physics. Large amount of traffic data is available from the Taiwan Highway System, where sensors are installed in highway segments for electronic toll collection. With the help of UROP students, we found interesting patterns of car flux and car density when congestion takes place last year. We will apply our knowledge to the prediction of traveling times under different traffic conditions using neural information processing methods.
If UROP applicants are more interested in fundamental aspects of neural computation, they can discuss with the supervisor about the individual topic. Possible areas include the use of deep neural networks to process images and fuse information from two channels, and the extraction of features in the intermediate layers of the neural network.
The applicants will have chance to process and analyze the traffic data obtained from the Taiwan Highway System, and other similar data sets, and use them to make predictions.
To extract meaningful descriptive parameters from large volume of data.
To use computational tools to solve real-life problems.