保研英语面试

Self-introduction(个人介绍)

Thank you for giving me this opportunity to interview with you. Now I will introduce myself briefly. 
My name is ***. I study at *** University majoring in computer science and technology(Innovation Class).
For the past three years, I have maintained to be top among students.
Along with my degree, I have participated in many contests. 
I have won more than ten provincial and national awards.
As for my research experience, I completed a paper as the first author and submitted it to the CCF journal IEEE Geoscience and Remote Sensing Letters, which is accepted. 
Through these experiences, I not only improved my skills in mathematical coding, but also gained a deeper understanding of research and innovation.
To enter your honorable school has been my dream for several years.
After the failure in college entrance examination 3 years ago, I didn’t give up this goal, and I firmly believe the effort is quiet worthy for this.
That's all, Thank you for your attention.

1. Why did you choose our school? Why choose this city?

你为什么选择我们学校?/为什么选择这座城市?

Firstly, what attrats me most is your resounding influence in the filed of Computer Science. 
Secondly, you have great scholars here, and I hope to get their guidence. 
Thirdly, the registered major has a certain correlation with my major. 
Finally, Nan Jing has great potiential for development, so it can give young people many opportunities for work and study. 
I think this is my ideal college.

2. If you can be successfully admitted to the postgraduate program of our university, what are your study plans for the next two or three years?

如果你能顺利的考入我校研究生，你对今后2年/3年的学习规划是什么?

Conduct more deep and systematic study of relevant courses, so as to lay a solid research foundation for the future. 
Widely read domestic and foreign literature to expand my vision in related fields.
Continuously improve my own scientific research ability, and actively participate in the tutor`s scientific research projects.

3. What are your strengths/weaknesses?

你的优点/缺点是什么？
我觉得我自己最大的优点就是坚持我想做的事情并且做事情有计划性。当我完成一件任务而且事情的结果又正合我的预想，我会觉得非常有成就感。但是我有时候感觉心理素质不是很强,这也是我在高考中发挥失常的原因而且自己有时候规划得有问题,出现抓不住重点问题。

I think my greatest strength is that I persist in doing what I want to do and do things in a planned way. 
When I complete a task and the outcome is exactly as I expected, I feel a great sense of achievement. 

However, sometimes I feel that my psychological quality is not very strong. 
This is also the reason why I didn't perform well in the college entrance examination. 
Besides, sometimes I have problems with my planning and fail to grasp the key points.

4. What do you think is the most important ability for students majoring in computer science?

你觉得对于计算机专业的学生而言，最重要的能力是什么？

I think code and maths are the most important abilities. 
When faced with a problem, we first have to build a model using mathematical knowledge and then solve it in code.

5. Introduce your hometown

介绍一下你的家乡

My hometown is DeZhou, a city in Shandong Province, eastern China. 
The city is known for its development in clean energy. 
One of its famous local food is DeZhou braised chicken, a well-known traditional dish across China. 
I love my hometown and I hope I can make some contributions to my hometown in the future. 

6. 你最喜欢的明星是谁？为什么?

My favorite star is Messi. He is a great football player.
When he was young, he was sick. But he worked hard and became the best.
I like Messi because he works hard and never gives up.

7. What is your best subject, and how did you apply it during your studies?

你最喜欢的课程是什么？为什么？

My best subject is Data Structures and Algorithms. I applied it by using C++ in algorithm competitions
, which improved my problem-solving skills.

8. Can you list some common data structures?

Common data structures include arrays and linked lists for linear storage, 
stacks and queues for order management, hash tables for fast lookup, 
and trees and graphs for network relationships.

9. What are the shortest path algorithms?

There are several shortest path algorithms. 
For example, Dijkstra’s algorithm works fast with non-negative edges, 
Bellman-Ford can deal with negative edges.

10. What are the spanning tree algorithms?

There are mainly two algorithms: Prim and Kruskal. 
Prim expands from one node, and Kruskal connects edges step by step.

11. Introduce your family

There are four people in my family, my father, my mother, me and my elder sister. 
My parents are strict with me, but they care about me very much. We are a happy family. 

12. What are your hobbies and interests?

I have several hobbies, such as football, ping-pong, and music. 
Sports make me healthy and focused, while music helps me relax. 
These hobbies make my life colorful.

13. How do you usually relieve stress?

你通常如何缓解压力？

For me, listening to music is one of the best ways to relieve pressure and let me free myself.
Also palying sports is another way.

14. 你读研究生的原因是什么？

The first point is that l don't think my studies during my undergraduate studies will make me competent for a job. 
The second point is that l would like to join the best school to do top-notch research and improve my own abilities.

12. 你为什么选择人工智能这个专业？

Because I think artificial intelligence is a very promising major, 
I hope to solve practical problems through the study of artificial intelligence.

14. 本科、研究生学习的不同在于？

本科生主要学习基础知识，而研究生则专注于研究和项目。

  Undergraduates mainly learn the basics,while postgraduates focus on research and projects. 

15. What are the innovative points of your XXX paper?

你XXX那篇论文的创新点是啥？
我们采用 HGNetV2 作为骨干网络，利用超图卷积和深度可分离卷积同时有效地提取局部和全局特征，在显著减少计算冗余的同时实现了高精度检测。我们设计了级联混合交叉头注意力模块（CHCAM），它融合了局部窗口自注意力和级联组注意力，逐层实现跨尺度特征聚合和自适应交互，从而增强了多尺度信息融合的能力。3）我们提出了四头自适应特征融合模块（FHAFF），它在传统的三头检测结构基础上增加了额外的深度特征提取头，以增强细粒度对象表示，并利用级联特征拼接实现全面的多尺度特征融合，从而加强跨尺度信息交互。

This project addresses the large-scale variation and complex characteristics of internal wave targets. 
Our paper's innovations are as follows:
1)HGNetV2 is used as the backbone, combining HyperGraph Convolution and depthwise separable convolution to
  extract local & global features efficiently, enabling accurate detection with lower computation.
2)We design CHCAM (Cascaded Hybrid Cross-head Attention Module), which fuses local window self-attention
  and group attention for cross-scale feature aggregation.
3)We propose FHAFF (Four-Head Adaptive Feature Fusion), extending the three-head structure with 
  a deep feature head, 
  enhancing fine-grained representation and multi-scale interaction.

16. What contributions did you make in your thesis?

你在论文中做了什么?

In this paper, I mainly worked on the technical innovation of the method. 
I designed the core algorithm, ran the experiments, 
and also revised and polished the final version of the paper.

17. Why are internal waves important to study?

Internal waves affect ocean mixing, energy transport, and even climate change,
so accurate detection is very important.

18. How does your model compare with existing methods?

It outperforms YOLOv5, YOLOv8, and Faster R-CNN in both accuracy and efficiency, 
while keeping low computational cost.

19. What challenges did you face during your research?

The main challenge is low contrast and irregular shapes of internal waves, 
which I solved using improved attention and fusion mechanisms.

20. What are your future research plans?

I plan to explore 3D observation and transformer-based architectures 
to further improve detection accuracy and robustness.

21. Why did you choose to improve the YOLO framework for internal wave detection?

YOLO’s lightweight and real-time advantages suit large-scale remote sensing data. 
Existing methods struggle with cross-scale feature fusion, so we enhanced it with 
HGNetV2, CHCAM, and FHAFF to better capture complex internal wave features.

22. How did you verify the effectiveness of your proposed model?

We conducted experiments on Sentinel-1 SAR and MODIS datasets, comparing with YOLOv3-tiny, 
Faster R-CNN, etc. CHGAFF-YOLO achieved higher F1-score, mAP, and maintained efficiency, 
proving its superiority.

23. What is your favorite programming language and why?

My favorite language is C++.
It has strong performance and rich features.
I like using the STL (Standard Template Library) — it helps me write clean and efficient code quickly.

24. How do you parctise programming?

I practice programming by solving problems online and joining contests. 
It helps me improve step by step.

27. Have you read any papers? Can you introduce one?

Introduce the innovative points of the papers you have read?

I read Attention Is All You Need. 
It proposes the Transformer, which uses self-attention instead of recurrence. 
This design is faster and more effective.

28. What is the difference between computer vision and image processing? Do you think they are related or different?

Image processing is about improving or changing images, while computer vision is about understanding what is 
in the image. They are different, but closely related.

29. What are the differences between C++ and python?

C++和python两者之间的区别在哪？

C++ is fast, good for system development. 
Python is simple and efficient, good for AI.

30. What is the difference between a process and a thread?

进程和线程的区别是什么?

A process is the unit of resource allocation, 
while a thread is the unit of scheduling and execution.

31. cnn和transformer对比 (A comparison between CNN and Transformer)

CNN uses convolution to extract local features, 
while Transformer uses self-attention to capture global dependencies.

32. 你是如何平衡竞赛和学业压力的?(How do you balance the pressure of competitions and studies?)

I balance them by setting priorities and managing time well. 
I focus on studies first and use free time for competitions.

33. 你感兴趣的研究?(What research are you interested in?)

I'm passionate about designing AI systems, 
especially in the fields of computer vision and large-scale models.

34. 对AI发展的看法?(What's your view on the development of AI?)

As a researcher, I see AI as a powerful tool, with the right guidance,
can push the boundaries of what's possible, 
but it must be developed with responsibility and fairness in mind.

35. Unet介绍

UNet is a type of neural network used mainly for image segmentation. 
It is named after its U-shaped structure. It has two parts: an encoder and a decoder. 

1）老师语速太快了，没听懂。
对不起，我不明白你的问题，你能再说一遍吗?
Sorry, l don't understand your question,Could you repeat it again?
2）听不懂老师提出的英文问题？
对不起，对于这个问题我并不清楚如何用英文回答。
Sorry, I don't know how to answer this question in English.
3）提问的专业问题用英语回答不上来。
对不起，我不知道。我对专业知识方面知之甚少。
但我希望我能有机会在您的指导下更深入地学习知识。
Sorry, I don't know. I am little acquainted with the aspect of special knowledge.
But I wish I can have a chance to learn more deeply the knowledge under your direction.