SOP

SOPとは

アメリカの大学院は大体statement of purposeという書類を提出する。 これは志望動機書みたいなもので出願書類で最も大事な物の一つである。これはかなり時間をかけて練って書いた方が良い。

型の重要性

僕は大学院出願を二年している。一年目は時間もノウハウもなかったので、とりあえず自分の思うように書いたが、とてつもなく酷いものが生成されていたと二年目にして知る。

こういうのは型があるし、読む側も型を想定して読む。なので型から外れた良い文章を書いたとしても全く評価されない。文学的な美しさは必要ない。必要なのは、型に当てはめて文章を書く力だけだ。

僕は大学の先輩や友人のstatement of purposeを参考に書いた。二、三人すでに海外大学院に進学している知り合いがいたので送ってもらって、添削してもらってと大いにお世話になった。

書くべきこと

SOPに書くべき内容は大体以下の通りである。

1. 自分の目標とその理由
2. エピソード1
3. エピソード2
4. エピソード3
5. エピソード4
6. エピソード5
7. 志望校の志望理由

僕が教わった方法はエピソードを5つくらい用意して時系列で並べる方式。各一つ一つに外部から証明できるような実績（degree, GPA, research paper, project link）などがあると良い。エピソード5は最終段落に繋がるように。エピソード1は最初の段落と繋がるようにして書く。

テンプレート

いかにテンプレートを貼っておく。これでMIT補欠合格を得たので、あり程度のqualityqualityは補償出来るはず。 実績のところ等は少し個人情報なので削っているがおおまかにはこれで良いと思う。

My ultimate goal is to contribute to the next generation of efficient computers. In order to push the computational capacity to the limit, I believe that it is essential to design computer hardware with its applications. To achieve this goal, I hope to enroll in a Ph.D. program and start my career as a researcher. In the graduate program, I plan to focus on developing deep learning accelerators.

To gain research experience as an undergraduate student, I visited two laboratories within The University of Tokyo to acquire research skills and learn about the important applications to be accelerated. In my first research project, (やったこと…) By visiting the laboratories and working on a consistent theme of acceleration of deep learning from different perspectives of computer systems, I realized that deep learning has various applications and that researchers in multiple CS disciplines are attempting to accelerate it.

With these experiences in mind, I decided to work on the acceleration of deep learning for my graduate thesis at (やったこと…). I felt that this would be an essential theme for applications that deal with significant decisions, such as in medical fields and aviation, which motivated me to work on the acceleration of Bayesian Neural Networks for my graduate thesis project.

I wrote my graduation thesis on Bayesian neural networks and presented it at an international conference, (やったこと, paperの内容) In these sessions, I discussed my research with professors and students from various countries and backgrounds, which motivated me to pursue research in such an environment, as well.

After writing my bachelor thesis, I looked back at my research and felt that I could have done better research in terms of social contribution. Specifically, I was thinking that I should know how deep learning is used in industry to conduct research that is truly beneficial to the world. I applied for an internship at (やったこと, internshipの内容). The fulfilling experience made me realize that we need more optimizations on algorithms and specialized hardware for deep learning, especially considering latency limitations, which will be an essential research field moving forward. In addition, latency requirements and how it should be optimized significantly depend on where it is used; thus, I have to consider more the use cases of deep learning.

Through my internship experience, I gained confidence that deep learning acceleration research is beneficial to society, and (internshipからの学び…) Therefore, I decided to work on training devices, focusing on the design of algorithms and hardware for training. I would like to continue working on this topic until I enter the Ph.D. program and ultimately acquire the knowledge and experience for my Ph.D. degree.

Massachusetts Institute of Technology (MIT) is my first choice for graduate studies. Without a doubt, the Electrical Engineering and Computer Science Ph.D. program at MIT offers one of the best ML and hardware research environments with the best professors in the world. Moreover, with so many excellent professors of diverse backgrounds at MIT, I hope to engage in collaborative research. In particular, the projects of XX Group led by Dr.XX are very attractive to me. The group is collaborating with Dr. YY to create an energy-efficient and low-latency deep learning accelerator. I found the tutorials published by the research group very intriguing, and I learned a lot from these tutorials when writing my graduate thesis. I had an opportunity to meet with Dr. XX and Dr. YY in August 2021 and reassured myself that we share similar research interests, and I felt that this was the research group I most wanted to join. Under the guidance of these great professors and in this multidisciplinary environment, I believe I can maximize my creativity to make significant contributions to computers. Through my education and research experience at MIT, I would love to see my future technology enliven societies and help people all over the world.