Neuromorphic Engineering: Reverse Engineering the Brain |
정두석 Neuromorphic engineering aims to reproduce several important brain functionalities that are realized by conventional computing at the cost of large computing power and thus energy. The functionalities of concern mostly include recognition and prediction. In my talk, I firstly address several remarkable feats of the brain in comparison with general-purpose computer, which markedly inspire neuromorphic engineering. Second, I overview the architecture of neuromorphic system from its building block to the entire system realizing complex neural topology. Third, addressed are the unique feats of dynamic neuromorphic system, distinguishing it from deep learning accelerators, and thus making it come into its own. Finally, I introduce the recent research activity of my group in the neuromorphic engineering field, which includes analog building block design, neuronal communication architecture, and fully digital neuromorphic processor named Neo2C neuromorphic processor. |