The NMC supports and encourages scientists to offer opportunities to students interested in gaining research experience. Over the years we have had many students researchers in such topics as high performance computing, general greenhouse work, administrative tasks, and plant biology research.
Students are paired up with NMC staff or researchers who need an extra hand, and this gives students the opportunity to gain valuable experience which will benefit them in their future careers. Benefits of summer work include gaining confidence, skills, and learning from mentors working in science.
The NMC would like to highlight one of our recent summer students, Charles Strauss, a Los Alamos High School student. Garrett Kenyon, a Los Alamos National Laboratory (LANL) scientist, hired Charles to work alongside LANL undergraduate student Jacob M. Springer, in a summer neuromorphic computing project. This work has led to a soon to be published paper titled, Classifiers Based on Deep Sparse Coding Architectures are Robust to Deep Learning Transferable Examples.
In this project, Charles and Jacob worked to test the robustness of one style of neural networks called sparse coding against dense coding, another style of neural network on adversarial examples. Adversarial examples are made by slightly altering an image, in order to cause a classifier to miss-classify. Charles states, “The alteration is so small that you can’t see it with your eye, yet the neural network goes from correctly classifying a picture of George Bush to thinking that he is actually Halle Berry.” The researchers found that sparse coded neural networks are more robust to adversarial examples than dense coded neural networks.
“Having robust classification systems that aren’t fooled easily is important”, says Charles. “If your Tesla decided that a fire truck was a good thing to crash into after encountering an adversarial example, you would wish your Tesla was immune to adversarial examples.”
Charles says he learned a lot and that this summer opportunity will strongly influence what he will choose to study in college. The NMC is proud of all the hard work Charles put into this project and congratulates him as being a co-author in this upcoming publication.
To read the about their work see: Classifiers Based on Deep Sparse Coding Architectures are Robust to Deep Learning Transferable Examples.