Tutorial II
10:50-11:30 AM, China Standard Time
Reinforcement Learning Methodologies for Package and Interconnect Design
Haeyeon Kim, Korea Advanced Institute of Science and Technology (KAIST)
Abstract -
This tutorial session introduces various reinforcement learning methodologies and their applications to specific package
and interconnect design problems. Recently, tremendous demands on higher performance, higher bandwidth/data rate,
smaller form factor, and lower power consumption are required for package and interconnect design. Due to the increasing
complexity of hardware design and the number of parameters to optimize, the application of machine learning has become inevitable.
Especially when it comes to signal integrity and power integrity (SI/PI) design for package and interconnect that involves extensive
simulations, reinforcement learning has shown promising performance in various tasks including decoupling capacitor placement,
routing, channel parameter optimization, ball map design, and equalizer design. In this tutorial, you may expect from the basic
mechanism of reinforcement learning to several applications of specific reinforcement learning methodologies for package and interconnect design.
Haeyeon Kim received her B.S. degree and M.S. degree in electrical engineering from Korea Advanced
Institute of Science and Technology (KAIST), Daejeon, South Korea, in 2020 and 2022. She is currently a Ph.D.
candidate in electrical engineering, KAIST, advised by professor Joungho Kim. Her field of research focuses on
machine learning (ML) application to power distribution network (PDN) design for 2.5-D/3-D ICs. She is currently
working on building neural architectures and learning schemes specifically designed for simulation-intensive signal
integrity/power integrity (SI/PI) problems.