|Keywords:||Energy-efficiency; Digital accelerator|
|Full text PDF:||http://hdl.handle.net/1853/58316|
System learning is the most fundamental research area in engineering domain. It is a modeling method to map external inputs to the corresponding outputs with/without physically analyzing the system between them. The system can be simple enough, e.g. a linear time-invariant system, to be easily identified by a simple mathematical model. However, it can be a more complex system, such as a nonlinear dynamic system, which is highly difficult to understand with mathematical representations. In this thesis, energy-efficient digital hardware to understand a wide range of complex systems using different approaches is presented. For a model-based approach, a programmable and efficient hardware for simulating dynamical systems is presented. The proposed platform accelerates the computation of solving a wide class of differential equations by utilizing a computing model called cellular nonlinear network with novel system architecture. As a data-driven approach, several neural network algorithms are selected for the system learning. The focused system is related to vision tasks such as image or video processing. Several design algorithms and analysis to realize low-power neural network accelerators are presented. The proposed low-power design methods are not limited to certain tasks, but are based on algorithmic analysis for general applicability.Advisors/Committee Members: Mukhopadhyay, Saibal (committee member), Yalamanchili, Sudhakar (committee member), Raychowdhury, Arijit (committee member), Kim, Hyesoon (committee member), Chai, Sek (committee member).