I have been working on my SoP for a while. And it would be very nice of you if you could help me to improve it.
Any advice is welcome! Thanks a lot!
When I first heard about Machine Learning (ML), I was skeptical that any machine could learn in a way that mirrors human learning. However, the online class, Introduction to Psychology, changed my perspective. I found that language is such an integral part of the human experience that it can be learned without any formal feedback or training. For example, even if parents did not speak to their babies or teach them how to talk, babies would still learn to understand language. It seems that language, unlike cooking or writing, is a completely intrinsic ability. This realization opened my eyes to the potential power of ML. ML gives machines innate abilities (preset models) to 'learn', just as humans 'learn' their mother tongue.
After that initial psychology course, I began my way up to Machine Learning. And I found that besides helping machines to become more intelligent, it can also shed light on human learning. For example, Deep Learning inspired me to think about how our brain might function, and the Bayesian Model helped me understand how knowledge can be summarized from observation. This duality arouses my interest the most, and is the driving force behind my wish to pursue a PhD in Machine Learning.
Also, I started to prepare myself for further ML research. Taking learning seriously, I have maintained my ranking among the top five percent students for three years. Moreover, I took the initiative to audit more related classes, including Convex Optimization and Functional Analysis. These courses provided me with a more acute and broader view on Machine Learning. At the same time, I continued conducting research based on my interests and obtained preliminary results. My hard and innovative work helped me win the Google Scholarship and the Guomo Ruo Scholarship, which are both highly prestigious awards and inspired me a lot.
To explore ML in more detail, I joined Dr. Xu's research group, where I attended weekly discussions for two years and came across fantastic technology such as Markov Chain Monte Carlo and Stochastic Gradient Descent. In addition, I made efforts to detect communities in social networks with attributes. Not satisfied with existing descriptions, which usually portray communities as nodes both densely connected and similar to each other, I proposed a new perspective to describe, detect, and analyze communities. The experiments I conducted demonstrate the accuracy of the proposed algorithm, securing results with about ten percent higher NMI, F-score and Jaccard Similarity over the baselines on four benchmark datasets. This work has been accepted by the 2015 IEEE International Conference on Data Mining (ICDM) as regular paper.
At the same time, I have continued to explore other areas of research outside of ML. For example, I led my school's operation system team in expanding our coursework into a project with a foundation of Ľ20,000. Our project aims to merge online spaces offered by different cloud storage services, in order to enhance privacy, speed, and volume by simultaneously uploading/downloading through different services. Those experience further confirmed my research interest in computer science.
Several months ago, I started an internship at Microsoft Research Asia. Through trying to learn Word Embedding on my own, I gained unique insights that also reveal the model's drawbacks. Thus, I am trying to enhance it with global information, through which I think the resulting model could distinguish words with similar usage but different semantic meaning, such as 'good' and 'bad'. Furthermore, I assist Dr. Ruihua, my mentor at Microsoft, in obtaining better predictions and perceptions of users' behavior on WeChat's Moments with topic models and neural networks. Recently, I started attending a reading group about Computer Graphics, which presented me numerous fascinating applications of Machine Learning technologies. Each piece of discussions I have had with researchers here has enhanced my understanding of ML, and encouraged me to seek out a graduate environment where I can continue this dialogue.
As I explored ML more deeply, I came across numerous insightful articles about Artificial Intelligence; and many of them come out of Stanford, like "WHAT IS AI". Also when I try to detect communities, I found the Stanford Network Analysis Project are both substantial and reliable. Furthermore, when I met problems and emailed doctoral students at Stanford, I was extremely impressed by their detailed, and supportive responses. This combination of innovation, and supportive culture underlies my desire to specifically attend Stanford, as I am certain it will provide me with the ideal environment for my doctoral work.
Aiming to be at the forefront of ML, I deem doctoral training critical for my career goals. I intend to gain deeper insight into ML as well as accumulate versatile experience across disciplines during my graduate studies. Though the choice of academia or industry can only be made after I have acquired a better understanding of both fields, one thing is for sure: I will continue with my research career upon graduation. Equipped with all the qualifications needed for the rigorous PhD program, I genuinely hope to inherit the spirit of Stanford and shape the future of technology.