Optimizing Mind

Deep Learning AI is powerful but much more difficult to create and update compared to the brain. Although there has been progress such as Transfer Learning, training still requires big data rehearsed in an unintuitive manner which makes debugging problems and implementation in real-life applications challenging. We developed an approach for much more efficient creation and editing of AI based on transfer learning up to two orders of magnitude faster using 1% of data samples while maintaining or exceeding current performance benchmarks. Our approach is supported by the National Science Foundation.