Winchester Thurston School
Winchester Thurston School
Teaching Computers to Learn
Nathanial Hull, William Lee-Moore, and Nathan Lovett-Genovese, Class of 2017
Class: Algorithm Design
Project: Teaching Computers to Learn
What it does: Computers are trained to find patterns necessary to solve problems through repeated exposure to sets of training data and training solutions.
How they did it: Using a graph with basic features like lines and circles, and thousands of random points, students input facts about each point. Once the net learned those facts, new data was presented, and it predicted facts for those points. In addition, Lee-Moore invented graphing software to display correct and incorrect data in different colors for better understanding of how the net was working and to pinpoint necessary changes.
Inspiration for project: While studying machine learning concepts, Lovett-Genovese discovered neural networks and was intrigued.
Target application: Speech recognition, medical diagnosis, and data mining are a few of the many ways in which neural nets are used.
Biggest challenge: Modeling curved shapes like parabolas and circles. When the problem— difficulty with the activation function—was finally solved, “There was whooping and clapping,” Hull shares. “One of the most exciting things about this was that solving the problem for that one implementation also solved the problem for any number of problems we might run through the neural net in the future.”
Importance of STEM at WT: “Projects with real-world applications give students the opportunity to become invested in problems around them,” says Lovett-Genovese. “Not only does this provide potential solutions to problems, it expands the horizons of students’ thinking to encompass the world.” Notes Hull: “Teamwork and communication skills are crucial in computer science, and projects like this force students to work together and learn that individuals must be able to agree on what they are programming and let the collective ideas of the group come before individual ideas.” Adds Lee-Moore, “Something that resonated with me about this project was the degree to which it was self-driven. We came up with and worked on our own ideas for potential future applications of the Neural Net; we took the initiative on developing software—the graphing program—beyond the initial scope of the project; and our teacher [Computer Science Department Chair David Nassar] always consulted with us about where the project was headed.
That drive, the investment in the project, a lot of it comes from a feeling of ownership and responsibility for the work. When teachers give students the responsibility to choose a challenge and then tackle it themselves, it’s a massive motivating factor in meeting the challenge.”
Valuable Lesson Learned: “I mainly reaffirmed the idea that I can get tough things done when I put my mind to it,” says Lee-Moore. For Lovett-Genovese, the project confirmed that “I really latch on to problems and don’t let go.” Hull’s takeaway: “I learned that I need to be more vocal and decisive about my ideas, especially when the group is at an impasse. I also learned that I am good at listening to other peoples’ ideas even during stressful and chaotic situations.”
