-
Notifications
You must be signed in to change notification settings - Fork 0
/
experience.tex
102 lines (87 loc) · 7.11 KB
/
experience.tex
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
% the mega teaching section
% \section{Honors and Awards}
% \cvline{2019}{\textbf{Novatek 2019 Scholarship}}
% \cvline{2019}{\textbf{Ministry of Science and Technology (Taiwan) Travel Grant}}
% \cvline{2018}{\textbf{Novatek 2018 Scholarship}}
% \cvline{2018}{\textbf{IEEE PerCom 2018 Best PhD Forum Presentation Award}}
% \cvline{2018}{\textbf{Principal Scholarship, NTHU}}
% \cvline{2017}{\textbf{APNOMS 2017 Student Travel Grant}}
% \cvline{2017}{\textbf{Principal Scholarship, NTHU}}
% \cvline{2016}{\textbf{Ministry of Science and Technology (Taiwan) Travel Grant}}
% \cvline{2016}{\textbf{ACM Multimedia 2016 Student Travel Grant}}
% \cvline{2016}{\textbf{Principal Scholarship, NTHU}}
% \cvline{2015}{\textbf{Pan Wen-Yuan Foundation Scholarship}}
% \cvline{2015}{\textbf{Ministry of Science and Technology (Taiwan) Travel Grant}}
% \cvline{2015}{\textbf{ACM Multimedia Systems 2015 Student Travel Grant}}
%\cvline{2011}{\textbf{National Tsing Hua University, Collegiate Programming Examination} {Qualified}}
%\cvline{2011}{\textbf{ACM-ICPC Asia HsinChu Regional Contest} {Honorable Mention}}
\section{Working Experience}
\entrybig
{\textbf{Networking and Multimedia System Lab, NTHU}}{Hsinchu, Taiwan}
{Research Assistant}{September 2019 - Present}
\innerlist{
\entry{Our research spans over the network and rich media.
We leverage various techniques to improve the quality of streaming applications,
e.g., 360\degree video streaming, VR cloud gaming, and 6DoF VR video streaming.}
\entry{Participated Projects: 6-DoF Immersive Video Streaming, and Machine Learning Platform}
}
\entrybig
{\textbf{Computer and Communication Center, NTHU}}{Hsinchu, Taiwan}
{Assistant System Administrator}{March 2020 - Present}
\innerlist{
\entry{We build and maintain the university-wide learning management website, which used by more than ten thousand students and faculty members.}
}
\section{Research Experience}
{\bf 6-DoF Immersive Video Streaming} {\it (Supported by the MOST Project: Teleporting Through Space
Across Time Using Head-Mounted Displays: A Case Study for Real Estate)}
Virtual Reality (VR) has become increasingly more popular in various business sectors.
The modern VR systems that support
six-degree-of-freedom (6-DoF) can provide more immersive experience, in which Head-Mounted-Display (HMD) user’s viewport can be changed
according to his/her position and orientation. However, because of the tremendous content size, 6-DoF immersive
video streaming dictates too much bandwidth and computing resources. In this work, we propose a configuration
optimizer that uses Reinforcement Learning (RL) and Convolutional Neural Network (CNN) to
select the best configuration setting. Through real experiments, we show that our solution reduces the bandwidth and computing resource consumption while delivering
good video quality.\\
\vspace{5mm} %5mm vertical space
{\bf Machine Learning Platform} {\it (Supported by the UMC Project: Development for AI Related Edge and Infrastructure)}
Machine Learning (ML) has been around for decades
and is now commonly used in many fields.
In recent years, more and more companies try to use ML techniques to achieve or improve their
productibility. However, capitalizing the potential of ML needs a lot of domain knowledge, along with tons of tuning for the best performance.
Furthermore, ML applications are not done after a model is trained. This is because the trained models may become outdated in the future, due to the drifts of concepts.
Therefore, after deploying an ML model, we still need to monitor its performance and retrain it whenever necessary.
% These problems lead to a requirement of software to allow developers to do these repeat jobs automatically.
To allow the ML developers to focus on analysis, we need an ML platform that can automate the routine tasks.
In this project, we build such an ML platform, which consists of various tools to speed up data preparation, model building, service serving, and performance monitoring of multiple ML applications.
We survey the existing platforms and generalize their components and functions.
This leads to a general ML platform design that can be adopted in diverse scenarios.
To demonstrate the practicality and efficiency of our design, we build a real testbed based on several open-source projects like Kubeflow.
We use the testbed to conduct a case study, which results in a few new research problems, that were not solved in the literature. We are currently solving these problems jointly with the UMC colleagues. \\
%\newpage
% \cventry{}{6-DoF Immersive Video Streaming}{(Supported by the MOST Project: Teleporting Through Space Across Time Using Head-Mounted Displays: A Case Study for Real Estate)}{}{}
% {Virtual Reality (VR) has become increasingly more popular in various business sectors.
% The modern VR systems that support
% six-degree-of-freedom (6-DoF) can provide more immersive experience, in which Head-Mounted-Display (HMD) user’s viewport can be changed
% according to his/her position and orientation. However, because of the tremendous content size, 6-DoF immersive
% video streaming dictates too much bandwidth and computing resources. In this work, we propose a configuration
% optimizer that uses Reinforcement Learning (RL) and Convolutional Neural Network (CNN) to
% select the best configuration setting. Through real experiments, we show that our solution reduces the bandwidth and computing resource consumption while delivering
% good video quality.
% } \\
% \cventry{}{Machine Learning Platform}{(Supported by the UMC Project: Development for AI Related Edge and Infrastructure)}{}{}
% {
% Machine Learning (ML) has been around for decades
% and is now commonly used in many fields.
% In recent years, more and more companies try to use ML techniques to achieve or improve their
% productibility. However, capitalizing the potential of ML needs a lot of domain knowledge, along with tons of tuning for the best performance.
% Furthermore, ML applications are not done after a model is trained. This is because the trained models may become outdated in the future, due to the drifts of concepts.
% Therefore, after deploying an ML model, we still need to monitor its performance and retrain it whenever necessary.
% % These problems lead to a requirement of software to allow developers to do these repeat jobs automatically.
% To allow the ML developers to focus on analysis, we need an ML platform that can automate the routine tasks.
% In this project, we build such an ML platform, which consists of various tools to speed up data preparation, model building, service serving, and performance monitoring of multiple ML applications.
% We survey the existing platforms and generalize their components and functions.
% This leads to a general ML platform design that can be adopted in diverse scenarios.
% To demonstrate the practicality and efficiency of our design, we build a real testbed based on several open-source projects like Kubeflow.
% We use the testbed to conduct a case study, which results in a few new research problems, that were not solved in the literature. We are currently solving these problems jointly with the UMC colleagues.
% }\\
%\newpage