The human brain has the remarkable ability to learn to recognize new objects from a single image. It achieves this by first training its neural networks on years of visual experience. In our proposed research, we exploit our knowledge of the biological central nervous system to design, implement, and train artificial neural networks that have similar one-shot learning performance. In Phase I we will demonstrate the validity of our unsupervised approach by training a deep attention network on artificial environments modeled and rendered with VTK and the Unity game engine.
