I hope better than my intern

LAU Tit rock is a Microsoft Research Asia Internet search and data mining group a research supervisor, devoted to the two aspects of the research work, one is through machine learning solution information retrieval problem (Learningto Rank), the second is the large scale data mining (Large-Scale GraphMining). By virtue of his university students who work during a wealth of experience in working shoulder shortly after the Group intern management. As a mentor, iron rock quickly identify oneself with the students, thus obtained a title of "kids". The following is a Research Institute interns and the "kids" dialogue. Author: Liu, University of science and doctoral

The "via" is occasionally inevitable

Q: I am from your personal home page that you have received a paper Communication and Image Journalof Visual Representation of the Most Cited PaperAward, is about the video content analysis.

This seems to be and you are currently engaged in information retrieval is not very relevant?

A: the video content analysis in Tsinghua Ph. period of study areas.

In fact I just entered in the 2003 Microsoft Research Asia, joining the group is also the media Computing Group at that time. In 2004, the Institute has been an internal restructuring, media Computing Group is split, so I transferred to the host's Internet search and data mining group, "to" start doing information retrieval research in this field.

From my personal point of view, chose information retrieval for two reasons: first, it is more responsive to current industry trends, you can make Internet era; the second is information retrieval of materials (such as text), and multimedia semantics of stronger than some, do not exist in the so-called semanticgap; because of this difference, engaged in information retrieval, we were able to put more focus on models, algorithms and theory level, instead of processing of the dimension.

Q: but this time "to" span is great, and today we look back, you retrieve the field with great success, and is in an extremely short period of time has been made.

You have the kind of experiences and expertise can you introduce?

A: I feel that reading the main harvest during Dr. lies in society research methodology (methodology), and not only applied to a field of expertise, so "to" did not think so difficult.

The main difficulty is how to get started faster. At this point, the first Institute in particular Internet search group gives us a good environment, a very high standard platform, by learning to colleagues and ask you a basic understanding of this area the most cutting-edge development. Secondly, we also need to know "what others are doing" or "we haven't done anything," found the missing parts to better locate my work. This learning process is very important, I brought together several interns started learning from theory. We spent six months to a year's time edge learning edge "lecture" information retrieval and machine learning related theory system brush. After we learned about you can be the entire set of innermost most understand theory (smiling). Learning forms, the most common is to talk, because it not only can you check yourself whether you have truly learned through this knowledge, but also help you with your problems and to others about the problem. It now seems that the six months to a year of the entry period is crucial.

Q: Please give us briefly about your current research interests, especially large scale figure of specific data mining refers to how a problem?

A: I am now studying direction mainly has two, one is through machine learning solution information retrieval ordering problems.

This is the information retrieval is very hot. We in this field, the results achieved are more leading, in academia has already had a big impact. Another is the large scale data mining. This figure refers to the network's hyperlink relations (hyper-links) form the scale of the graph, a rough estimate only China area contains approximately 10 to 20 million nodes, there are several tens nodes worldwide. Analysis of this massive data is quite challenging, whether on computing resources or on the theoretical analysis have challenges. Previously through the analysis of small figure got some conclusions; these conclusions in large scale can also apply to would appear from changeing process, these are not the answer. We are ready to build a system that we call Graphor, with the help of distributed computing platform to handle such a large scale of the map, look forward to derive meaningful conclusions.

I intern says: I hope you are better than me

Q: I am from your interns receive some information.

They said to you on their instructions very carefully. Your trainee project?

A: I've just joined the Internet searching for groups, group leaders mawei ying, just let me in charge of our group of interns work, such as the Organization of the seminar, project review, BestRank Camp, winedown etc, I had about 2 years.

In fact, be able to provide help and guidance of trainees is a great honour. At the same time, I put the guide their interns with a high level of academic research as an important, evenThat is the key for a given period. I was holding a very sincere and serious attitude to treat each student to Research Institute internship. In fact, the intern's supervisor send them to the Institute, is the hope that they can here be greater progress; I can say for sure, the Institute will not give interns as "labour"; we really want to be and domestic universities, research institutes to establish long-term relations of cooperation. For example, I was at the time of Tsinghua University professor tutor Zhang Xudong and CAs academician of Mathematics by Franco were exhibited their doctoral candidates rushed over long-term internship, we made a lot of very good results.

As far as I am concerned, I have my internship is very emotional.

Several early I intern, for example, now also in the practice of qintao, and has become an employee of Gao Fang Wu, and I start doing research this subject. I think we are growing together, with progress. I am Dr. "soil", due to the limitations of our laboratory, I read the doctor's time has not yet come into contact with the international top-level research results, or even not-so-familiar with foreign counterparts who have done very well. I hope that my students were given to the Institute, since there is such a good one platform, we can make better, more excellent than I can.

Q: I know your interns have many projects.

You will guide me of your students, hands-on teaching them to solve the problem. But I also heard about another way of saying that this guidance will allow them to develop a dependency, leaving you to help yourself to do research. What do you think?

A: I need to explain.

I started doing research for students, for example, prepare a top-level conference papers, I did on his guidance is very good. From the topic, I'll help him; to do the experiment, we will discuss, I would help him solve a specific problem; to finally write a thesis, I'll help him instead, or even changed many times. These are very meticulous services, but the purpose is not to replace the students do, but to let the students familiar with and understand the entire process of doing research. Clever students during this process already can learn what their future independently engaged in academic research. Wait until the second, third, I am not going to help so many busy, but mainly to provide direction and guidance on key technology issues, will make his own complete more or even all of the thesis work. I will not let my students, not brains alone I help do the research. From a practical point of view, I am thinking of interns are also very active, guts, brainstorming, speaking always positive, enthusiastic "shoot brick", and even some employees describe them as "arrogant". I feel very good, I encourage them to head in authoritative.