A neural data structure for novelty detection

This work is significant for two reasons. First, considering fly olfaction in the context of a computational algorithm emphasizes features of olfactory processing that might otherwise be less apparent. For example, comparison of prior approaches for novelty detection makes clear that different odors are not just novel vs. familiar, but rather can have various degrees of novelty depending on the similarities between odors. Second, understanding how the fruit fly olfactory circuit detects novel odors can inspire new methods for similar problems in machine learning. In this way, both computer science and neuroscience benefit from the comparison of these two systems.