The reason why I decided to further my postgraduate study at the NUS is that I genuinely felt the strong academic atmosphere and Professors’ profound academic skills of this institute when I participated in the summer workshop in 2019. At that time, my tutor professor Kok-Lim Low, was willing to discuss with me based on my ideas and continuously helped me in the final project. Without his enthusiastic guidance, it would be impossible for my work to achieve such a high ranking in the end. From then on, I was deeply fascinated by its remarkable teaching quality. Besides, my passion for intelligent systems learning is rooted in several types of research in this realm during my undergraduate education in computer science and technology. The two most impressive experiences were the AI learning of classic algorithms in reinforcement learning (such as the Q-learning algorithm), and the research on speech enhancement while participating in the only speech-related project of the school. Currently, the city Shenzhen I am living in is undergoing significant changes, and the application of artificial intelligence is indispensable. As such, I yearn to devote myself to this revolution. Therefore, I need to receive more systematic and further education in this area so as to meet more challenges.

Why Program

The Master of Technology in Intelligent Systems program at the NUS is very appealing to me because it’s highly compatible with my career path and meets my desire. With an expertly designed curriculum, it focuses on the holistic learning of theories, techniques, and methods of artificial intelligence. Especially, I am especially intrigued about modules like Pattern Recognition Systems, as previously I have researched speech enhancement in the only speech-related project of the school. From where I stand, I think that speech enhancement is of considerable significance to the accuracy and efficiency optimization of pattern recognition. Through learning this course, I hope that I can apply the confrontation network Generative Adversarial Networks to speech enhancement and pattern recognition, and I believe that this deep learning network of GAN can be widely applied in the field of speech, and shine in the future. Combining these factors, I deem studying this MSc course will be the best choice for me to extend my previous studies and prepare for my career.

Academic Experience 学术经历，相关课程

Applying to such a highly demanding university means facing off against intense academic competition, but I firmly believe that my record stands up to scrutiny. Through constant efforts, I have underpinned the rudimentary knowledge and skills of programming, operating system, computer network, data structure and algorithm in CS. Notably, the excellent grade in Fundamentals of Programming and Artificial Intelligence I have achieved demonstrates my good mastery of C++ and Python. Studying Principles of Data Base did not only equip me with knowledge of the basic structure and characteristics of the database, but it also honed my logical thinking and hands-on skills. On top of that, aware of the significance of mathematics, I have also attached much importance to learning linear algebra, probability theory, and calculus, for gaining a strong mathematical and statistical foundation to succeed in your course. I think that the theoretical knowledge I have acquired, no matter in aspects of computing or programming, will be a cornerstone for me to accept more demanding learning.

研究经历

I’d like to continuously surpass myself and hoist my research and analysis skills in both academic study and scientific research alike. Taking the research on reinforcement learning in the field of AI as an example, I determined and grasped the research algorithm of Q-learning within a limited one week, and successfully designed a path-finding robot that can automatically learn to obtain the shortest path of the maze. In the initial three days, I finished the basic programming. However, our group encountered one problem the training process of visualization needed much time to reach for some specific maze situations. To handle this, I analysed it from two perspectives the parameter (learning rate and decay rate) setting and programming code. The adjustment of parameters reduced the training time but not was conspicuous. Then through modifying the codes controlling the boundary of stopping training, the training speed was much improved. Our joint efforts finally paid off as our work was recognized as excellent by the teacher. Aside from strengthening my knowledge and skills in showing the algorithm in a concrete visual way, I also further unearthed my potential in compiling programs with a comprehensive application of multi-domain expertise. Another fact I would like to mention is that I am proud of the algorithm skills have developed so far. I won unanimous recognition from professors and assistants as I was the only one in the class using a batch of Python packages including Matplotlib, NumPy, and Pandas to analyze and visualize the performance of the algorithm under different data volumes and data types. These abilities, I believe, will stand me in good stead for advanced learning.

交换经历

Knowing the importance of combining theories with practice, I joined the NUS School of Computing Summer Workshop in 2019, which only accepted 12 students around 180 people. There I selected real-time rendering as a research topic and conducted our work based on OpenGL shader language. Three single tasks, two group tasks, and a final report are covered in this project. In the early stage, everything went on smoothly. However, our team was uneasy about progressing when tasks involving in more difficult algorithms knowledge such as bloom special effects, texture mapping, and ray tracing. So I took on a leadership position, organizing a series of seminars to ensure everyone acquires in-depth understanding. It worked out. In the final process, each member shared the knowledge mastered and effectively filled the knowledge gaps of other members. As a team leader, I often stayed up late, checking the code to eliminate errors to ensure smooth coding. Eventually, we successfully designed a continuous ball motion animation of the scene that can be automatically converted according to the timer in Shadertoy. This experience was unforgettable, as it not only enhanced my programming ability and deepened my understanding of the knowledge learned in class, but it also strengthened my leadership and problem-solving competence.

career path 职业规划

With regarding my career goal, I aspire to hunt for the first job in the department of natural language processing and speech enhancement at Microsoft Research Asia. In the short term, I desire to continue to propose a new speech enhancement algorithm suitable for natural language processing. The speech field can be said to have just started, and new algorithms and models are widely required. Thinking longer-term, I hope to return to my alma mater, aiming to apply speech enhancement and recognition to develop an underwater robot that performs well in the deep sea, and can complete survey functions in the deep dark sea through sound signals; thus contributing to China's deep-sea undertakings. With such plans, I look forward to enrolling in your program.