An awardee of the Career Award from National Science Foundation’s Faculty Early Career Development Program, Paul Hand had to investigate the algorithms for new deep neural networks, which are able to use machine learning techniques to construct useable images from significantly fewer data. For Hand, who is an assistant professor at the Northeastern University, these algorithms look like puzzles begging to be solved. He says, “We don’t have good justifications for why these neural networks tend to work”.
The new algorithms could mean that the MRI patients will have to spend less time in a claustrophobic tube. In earlier computer vision techniques, researchers would tell algorithms which aspects of an image were important. If the researchers wanted an algorithm to identify pictures of a pig, for example, they would need to define the characteristics that make a pig different from a building or a glass of water.
But it turns out that humans aren’t terribly good at choosing which characteristics a machine needs to have at its disposal to discern the difference. The machines do better without our help. Deep neural networks are designed to teach themselves which characteristics are important for their particular task. Researchers train neural networks by providing them with tonnes of data to practice with. Each time the network produces the right answer, whether it is accurately reconstructing an MRI image or pointing out a pig, it learns. With enough, varied, training data, the network can figure out which combination of edges and corners makes the shape of a pig’s face, and that pig faces are important for finding pigs.
Hand describes, “It’s very much like you’re in math class, and you have homework every day for the rest of your life. But you have to come up with the homework problem yourself, and it might not even be solvable. The thrill of the profession is that you get harder and harder puzzles.” They are trying to bring in theory and principles into recovering and processing images with neural nets. They are really living between these two fields of computer science and applied mathematics.
Radhika Boruah