Greeting from Taiwan! Good evening : )
I am intending to apply the admission for KAIST graduate school of artificial intelligence via Korean Government Scholarship program (KGSP).
The following content is my personal statement for the program.
KGSP Personal StatementWe are assigned to cover:
- Motivations with which you apply for this program
- Family and Education background
- Significant experiences you have had; risks you have taken and achievements you have made, persons or events that have had a significant influence on you
- Extracurricular activities such as club activities, community service activities or work experiences
- If applicable, describe awards you have received, publications you have made, or skills you have acquired, etc.
Please give your thoughts and advice. Thank you very much!
It was Christmas eve, I had my neural network model looking at the deep starry sky through the best space telescopes in the world, recognizing tens of thousands of supermassive black holes hosting at the center of distant galaxies. I swept quite a few coffee and Redbull cans from my desk to the trash can, realizing I just made a significant improvement on a notoriously difficult cosmology problem with AI technique.
This was the day I made up my mind to study artificial intelligence-I showed myself how AI could, and gonna, change the way we treat problems. This experience motivated me to join the ranks of the scientists who improve AI and apply AI in our civilization, and it is why I am motivated to apply for this program.
I graduated from Taipei Municipal Jianguo High School, a high school which requires the highest scholastic performance in Taiwan. I am currently an undergraduate student majoring in physics at National Tsing Hua University (NTHU), which is normally considered as one of the best universities in Taiwan. In addition, I had been an exchange student at Seoul National University (SNU) for a year. My parents are both civil servants, engaging in the Ministry of Justice and Ministry of Health and Welfare in Taiwan. In addition, I have a grandmother and a younger sister in my family.
I have the necessary background to conduct researches in the field of artificial intelligence for the following reasons: (i) I have published a paper based on neural networks in one of the world's leading astronomy journals as the first and corresponding author. (ii) I have worked as an intern in Kneron Inc., an international AI enterprise, and have made contributions to the improvement of the commercialized face recognition model. (iii) When I was an exchange student at SNU, I once proposed and wrote a thesis about a novel architecture based on double deep Q-learning (DDQN), showing my idea improves the performance of DDQN.I once summarized my research as a paper, conquered the peer-review process, and published the paper in the journal.
As a physics major student in NTHU, I joined Prof. Tomo Goto's cosmology group and sought for my undergraduate research project. After quite a few literature surveys, I noticed that recognizing active galactic nuclei, a type of galaxy cores powered by supermassive black holes, is a problem that remains unsolved. Instead of the classic statistical analysis method, I came up with the idea that using a neural net to do the classification might be more effective. As a result, my method based on convolutional neural networks exhibits a significant improvement in performance compared with the traditional method. Finally, my paper "An Active Galactic Nucleus Recognition Model based on Deep Neural Network"
is accepted and published in one of the top astrophysics journals "Monthly Notice of Royal Astronomy Society (MNRAS),"
and I am the first author and corresponding author of it.I was hired as an intern in Kneron Inc., an international enterprise providing edge AI solutions, and this experience has strengthened my ability in developing AI-related applications.
I submitted my CV to Kneron, got the intern offer, and cumulatively worked there for more than 6 months. During my intern period, I was responsible for improving the performance of Kneron's face recognition model. My mentor Mr. Vincent Shen assigned me to build an application that could develop a face database, including images and labels, in an efficient manner. I make the application features with customized model selection, embedding auto-clustering, MongoDB connection, GUI interface, etc. Eventually, Kneron has hired several workers using this application to build a face database owned by Kneron.I wrote a thesis about an improved architecture regarding deep double Q-learning (DDQN).
When I was exchanging at SNU, I took a deep reinforcement learning (DRL) course and was required to complete a thesis at the end of the semester. We know that deep Q-learning (DQN) always suffers from the overestimation problem, thus previous studies have proposed architectures like DDQN to avoid it. In my thesis, I proposed MDQN, a modified architecture based on DDQN but train and displace a set of online networks sequentially, with the target networks relaying between the displacements. The test results on an ATARI game show my MDQN is likely to be better than the popular DDQN. Finally, I got a score of A0 in the DRL course.
In summary, I have an SCI journal paper about AI application for cosmology research, an intern experience in an international AI enterprise, and a thesis regarding an improved DDQN architecture. I am looking forward to advancing my knowledge of AI in your program.