Biography
My name is Jin-Siang Lin, a graduate student with Computer Science background.
Currently, I am working at Synopys as a R&D.
I'm responsible for developing API for Verdi debug tool.
Email
Github
Linkedin
CV
Work Experience
Synopys, Hsinchu, Taiwan
Verdi R&D
Date : 11/2022 - now
Responsible for API development for Verdi debug tool.
MediaTek, Hsinchu, Taiwan
GPU software engineer
Date : 09/2021 - 10/2022
GPU driver and application development, especially on Vulkan API.
Responsible for Vulkan layers and tools development for GPU performance tuning.
Qualcomm, Hsinchu, Taiwan
Machine learning engineer (intern)
Date : 04/2021 - 07/2021
Responsible for deployment of machine learning flow for semiconductor manufacturing data.
Optimize the original data-pipeline for 43x faster.
Research Experience
Policy Transfer / Policy Composition on Reinforcement learning
Date : 10/2019 - 08/2020
Our research focuses on the policy transfer on robotic simulation and application
We come up with a good way to decompose the behaviors of policy on pre-trained tasks into some sub-primitive
policies
We further leverage these sub-primitives to increase the sample efficiency on transfer tasks
Medical image colorization
Date : 08/2020 - 03/2021
My new project focus on Medcial image colorization (e.g. CT image) using Deep Learning methodology.
Publication
Hierarchical Alternative Training for Long Range Policy Transfer
ICML 2020 BIG workshop
Second Author
We proposed a framework called Hierarchical Alternative Training(HAT) that leverages the hierarchical
structure to train the combination function and adapt the primitive polices alternatively, to efficiently
produce a range of complex
behaviors on challenging new tasks.
Toward Robust Long Range Policy Transfer
AAAI 2021
Second Author
We propose a method, which leverages the hierarchical structure to train the combination function and adapt the
set of diverse primitive polices alternatively, to efficiently produce a range of complex behaviors on
challenging new tasks.
