TinyML squeezes deep learning into the resource-constrained IoT devices (microcontrollers (MCUs)) and becomes promising in multiple application domains such as smart homes, smart cities, and intelligent manufacturing. In addition, TinyML offers a cost-effective, low-power, and high-privacy solution for current machine learning applications. In this talk, I will present the latest trends in hardware and software solutions for TinyML models. In addition, I will address the challenges of deploying TinyML models to AIoT devices. Finally, I will introduce deploying TinyML to the AIoT MCU device.

葉宗泰助理教授於2020年在普渡大學獲得電機與計算機工程博士學位，指導教授是Timothy G. Rogers。在國立清華大學資訊系統與應用研究所取得碩士學位。之前，曾在AMD研究部門、普渡大學和中央研究院工作。研究成果曾被提名為最佳論文獎（PPoPP 2017），並且曾獲得普渡大學的Lynn獎學金。

As computer-based systems demand more and more performance and functionality as well as better programmability, the derived sophisticated parallelism and speculation requirements have continued widen verification effort gaps in signoff and yet adding opportunities to creativities to design verification engineers. Nonetheless methodologies ordinarily used are often superficial due to immature theoretic and practical foundations in confining complexities for essential scalabilities, which makes related knowledges, methodologies, or skillsets needed for driving scalabilities super precious. In this presentation overview of modern processor verification concerns and evolutions of methodologies would be explored from the aspect of professional value.

Chi-Ming Li is a verification engineer at Synopsys, prior to which he worked for Arm. He received a B.S. and a M.S. from NTHU EE. Chi-Ming is passionate about computing technology and has hands-on experience in several processor verification projects.