When I was fascinated by the world of computer science, I was taking the second grade of primary school and was recommended to take the Pascal programming competition training program. Though my duty was to solve the mathematical problem via coding, I found pure joy in the process of using an algorithm to calculate the answers to questions that is hard to be calculated manually. This course showed me the magic power of computers and stimulated my further efforts to use the computer to resolve problems in the real world. Therefore, I selected computer science as both my undergraduate and graduate field. An intensive study and research experience during the graduate period stoked my passion for interdisciplinary subjects such as artificial intelligence, machine learning, data mining, medical science and brain-computer-interfaces, etc. With the development of computer technologies, many valuable systems and products have already transformed daily lives and made our life more exciting, efficient, and intelligent. Considering the significance and my pursuit for advanced knowledge, I make up my mind to pursue a Ph.D. in Computer Science to obtain state-of-the-art science so that I will be able to devote myself to contributing to the further realization of splendid technologies.
During the undergraduate stage, I have established a solid foundation in mathematics and computer science, which provided me with a good foundation for pursuing advanced study and research. Further, my interest in engaging in the research project has motivated me to take some projects which allowed me to apply theoretical knowledge in practice to extend my hands-on ability and research qualities. For example, based on HTML5, I used the AppCan framework to write search modules and design an educational APP user interface to help students comprehend the homework better; I built an R language computing cluster of the high-performance-computing center of the department based on Spark to speed up the calculation; I used C# language and WPF framework to write the upper computer software of the satellite link simulation chip to visualize equipment status information; I wrote popular game ‘FlappyBird’ in Java for fun. In addition, I have successfully got two software copyrights. In the spirit of open source, I built a personal website via the Hexo framework and shared dozens of project codes and scientific research experience on it, which gives me a chance to communicate with others.
Encouraged by Geoffrey Hinton and Terrence Sejnowski, great computer scientist and cognitive psychologist, I decided to learn cognitive neuroscience and artificial intelligence during my graduate stage. Satisfactorily, I found an excellent graduate advisor Dr. Zheng Li who worked on BMIs at the State Key Laboratory of Cognitive Neuroscience and Learning of Beijing Normal University. Since then, I have been working on the research of Decoding Algorithms of Motor Cortical Brain-Computer-Interfaces Based on Phase-of-firing Encoding. BMI systems translate neural activity from the brain into control signals for prosthetic devices, which assist people with paralysis by restoring lost function. Before I was admitted to graduate school, I was not unfamiliar with the brain, animal experiments, not to mention doing craniotomy on my own because of the lack of background knowledge. Step by step, through learning and practice, I found that the research of neural engineering is a complex combination of mathematics, statistical signal processing, machine learning, low-power circuits and real-time system modeling and implementation, etc. It is very challenging but makes me excited. To design animal experiments, the knowledge of psychological experiment design and neuroscientific background knowledge is required. Training monkeys to control the joystick for finishing appointed behaviors requires personal qualities such as physical strength, endurance, intelligence, and courage. And, the configuration and maintenance of experimental equipment test researchers’ knowledge of low-power circuits and mechanical skills. Besides, surgical operations such as drilling the skull and headpost implantation require the operator’s circumspection and cooperation. Furthermore, conducting neural signal analysis including neural decoding requires the integrated application of abstruse mathematical knowledge, modern machine learning theories, and strong programming ability.
To keep up with the quick development of state-of-art computer science and follow my interest in machine learning, I also engaged in some practical projects. Taking the project of Explore which customer characteristics are meaningful for predicting cellphone customer loss based on SVM with R language as an example, I deeply applied some statistical model to do prediction. Specifically speaking, with a tuned Support Vector Machine, I implemented data classification and visualized the SVM model. Based on the K-Fold Cross-Validation method, the model performance was compared with other classification models. Finally, the prediction accuracy reached 93.12% after optimizing the parameters. Besides, I did handwritten digits (MNIST dataset) recognition based on denoising convolutional autoencoder. By doing projects, I was familiar with the process of data mining, machine learning and deep learning process, including problem analysis, data exploration, data preprocessing, feature construction, feature selection, model selection, model optimization, model fusion, model comparison, etc.
Currently, I am still participating in two projects. One is for testing a Low-intensity Transcranial Ultrasound Stimulation device on monkeys, I conducted monkey’s brain MRI scanning and utilized FSL-fast and BrainSight to perform segmentation of the 3D reconstruction of the monkey’s skull. Another is ‘Cognitive State Assessment and Application Technology’, I offered some suggestions on experimental design, helped select appropriate ImageNet pictures for fMRI tasks and conducted an fMRI data acquisition which includes guiding the subjects to do visual tasks and fMRI scanning. In the next few months, I will be involving in collect MEG/EEG data from subjects for cognitive tasks and decode the people’s cognitive state of EEG signals.
In recent years, deep learning has brought new breakthroughs to artificial intelligence, but artificial neural networks and deep learning only learn part of the topological structure between neurons and are still in the initial stage of brain simulation. There is still a lot of information in the structure and function of the brain that has not been discovered, which contains huge possibilities. Prof. Geoffrey Hinton, the winner of the ACM Turing Award, believes that the key to overcoming the limitations of artificial intelligence is to build ‘a bridge between computer science and biology’. As neuroscience continues to deepen and accumulate an understanding of the human brain, brain science and artificial intelligence research have become closer and closer. The two fields learn from each other and merge, which is expected to promote a new era of the intelligent technology revolution. I hope my interdisciplinary background in computer science and cognitive neuroscience will help me to propose new ideas in scientific research.