Friday, May 27, 2016

Discerning the Forest from the Trees - The Insights from my PhD Supervisor JC Browne

In the acknowledgement section of my PhD dissertation in 1994, I wrote:
I appreciate the support and help of my advisor, Dr. James. C. Browne. His vision and ability to abstract away the details has taught me how to discern the forest from the trees. Working with him, I learned the full import of what it means to ask the right questions. His questions would often send me scurrying on a search path that would clarify my confusions and would lead me to a true understanding of the problem, and hence, to the solution.
As I contemplate on the above statement written over 22 years ago, I think I need to explain the full import of this insight from my five years of effort for PhD at UT Austin from end of 1989 to the end of 1994.

James C Browne, my Supervisor (Profile at the end of post)

But, first some interesting background. After completing my MS at Austin, I was searching for a supervisor for my PhD who can sponsor my research assistant-ship. At Austin, RA ship had to be sponsored by the supervisors from their funded projects. Some supervisors were able to obtain huge research funding while others fed upon their projects. I learned that JC was among the largest grant-getters at UT Austin, and had a huge "empire" consisting of several research labs, scores of research students and several associated faculty members. His students were from electrical and computer engineering department, computer science department, aerospace department and also some other departments as described in his profile at the end as Browne received the 2004 University of Texas at Austin Career Research Excellence Award for maintaining a superior research program across multiple fields over a 45-year career.

Rigor of Graduate Courses


How I ended up joining the parallel programming group of CODE and chose him as an advisor is an interesting story. Someone had told me that he offered a course in which everyone used to get an A. The last piece of information was of great interest to us who were always tempted by the short-cut. On verifying that it was the case that for the past several years every student of JC's graduate course had always got an A, we promptly enrolled in his course. His course was offered at 3pm. I used to come to the class after consuming a sumptuous lunch with parathas and dahi with obvious results: The moment JC would turn off the lights and start flipping from the tons of transparencies that he would normally show in each session, I would doze off and take my afternoon siesta. The major graded part of the course required a project of programming on 3 parallel computing machines using 3 different parallel programming languages a parallel algorithm published in a paper. We were 3 people in the group, and each was given a different parallel algorithm to program on these three machines. We were given a framework on which to report the results of programming. What this course turned out was a nightmare. Programming in 1990 on three machines over the internet of that time using ftp and without the use of hyperlinks and online resources was a tedious process. With scant published support on the programming languages and the lack of support for the operating systems of the experimental machines and trying to program them over the network was a very fragile process where the connections kept on breaking. But, we trudged along. I think I ended up working so much that I had not worked in any course before or since. In the end, the A's that all of the students were well deserved given more for our dogged determination in making our algorithms run on those fragile machines and rickety networks through our sustained effort.


How to Select Supervisor


After I had joined the CODE Visual Parallel Programming Research Group, one of the group members who was a Post Doc working on the project asked me why had I taken JC as my supervisor? Didn't I know that the research students of JC typically took at least 5 years for their PhD!  But, this piece of information came too late. I had already committed and I could not wriggle out from my commitment. I realized the hard way why they took so long. But, I am so much the better for that.

His style of supervision was that of  Laissez-Faire where he let us do what we want and would give us as much latitude as we want, eventually building up the pressure from our own side that we need to complete it asap, because he will not ask us to hurry up for another ten years! He would let us be on our own for months at end without even asking us to report! Something which is unheard of for younger supervisors. During the initial period I took a lot of time going over the related literature. Read as much material I could find and tried to identify the gap on which to work on. This was at times really boring but in the end quite rewarding.

I learned it the hard way that if if you want to complete your PhD early, select a supervisor who is hungry for publishing research papers and acquiring tenure track. He will sit with you, edit your work, even work on derivations with you and would some how make sure that your work gets published as soon as possible. If, however, you select a senior supervisor who is not as much hungry for publishing papers, you can still learn a great deal from him, but you would have to do it mostly based upon your own internal motivation and will. Extracting supervision from a senior and a busy faculty member requires your commitment and continuous persuasion and dogged determination.

How to identify the weakness in research writing


Often I would delay going to him for months, waiting till my work was flawless. Eventually when the pressure built up from within me and I would prepare a progress report for submitting to him, I would try my best to hide away the major problem from him in the report. Typically, the progress report would consist of tens of pages. As I would wait for him outside the door to call me in his office a few minutes before the penultimate time, I  often noticed that he would pick up my report a couple of minutes before my time and would just turn the pages skimming over them. At the appointed time, the moment I would settle in my chair, the first thing he would often ask was the question pertaining to the thing that I was trying so hard to hide from him. When I got my PhD and went for the last time to take his leave, I could not resist and had to ask him as to how he could identify the problem buried deep in my report. The reply he gave me still resonates in my ear:
"Dr Hyder [the first time he addressed me as that] several years from now when you would have supervised so many theses as I had done, you would also know how this gets done".
How true. Only now I am beginning to discover what he meant then.

Thinking abut how those questions actually clarified the difficult knots for the past so many years had led me to the following conclusion: These questions emerged, I now tend to believe, not as much from the reading of the write up of the student, but more from that deep knowledge and gut feeling that comes from following a field for a long time, and realizing where the field is headed in the long term and what the impediments on its path are. His questions would connect the research to those questions that he had been thinking about, and the researcher is grappling with them "implicitly" but is only made aware of them "explicitly" through the question.

There are very few new ideas in this world


One day I went to his office and there was no empty chair to sit on. There always used to be piles and piles of old and newly arrived issues of journals and conference proceedings. It looked as if all the journal editors and conference editors are his friends or indebted to him or have his subscription and are thus sending all their issues regularly. That day, every chair and every nook of the office seemed to have these piles of journals issues and conferences proceedings. He stood up, went to a visitor chair, lifted the pile lying on that chair, making space for me to sit, and placed the lifted pile elsewhere on the floor. Then pointing to these piles of research papers, he said what haunts me to this day:
"My Hyder, you see all these research papers. There are very few new ideas in them! Mostly they are the rehashing of the old stuff, only in a new terminology"

Mathematical Representations and Reality


I got a lesson in this the hard way. As I was grappling to identify a mathematical formulation for the idea on which I was working. I had started with simple set theory formulations based on first order logic. Then, I came across a paper that had used Temporal Logic. I became fascinated with it and asked JC about it. He said that he had a very good book on Temporal Logic, which he, then, loaned to me. I spent the next 3 months trying to understand the book and then to implement my work using Temporal Logic. Instead of making my life easier, I found that Temporal Logic seemed to be making my expression more difficult. I went to JC again telling him that I think Temporal Logic will not work. But I quickly added:  In the mean time, I have discovered another field, called category theory and morphisms. Then, I asked him should I try using this? He said OK. Go ahead and try it. I went back and spent a few months working on that. Eventually, I returned to him empty handed without having made much headway. Then he gave me a lesson that I would never had understood, had he told me in the very beginning not to go on this wild goose chase. He said:
"Mr Hyder, you see all these mathematical systems are just trying to represent the reality. There is no technology in the system, only representation. I think you can achieve what you are trying to do with the simple First Order Logic that you were trying to use earlier. Why don't you continue the work from where you left off!"

I was so much relieved. The whole burden had gone away and I had learned an important lesson. Mathematics is not some thing to impress people with in your paper. You need to use the simplest of representation that meets the requirement. Fancy mathematics does not add any power to your basic idea!

For Supervisors: I think it is part of the job of the supervisor, to allow the research students to use their hunch, allow them to make mistakes, and learn to readjust the hard way; but during all this exploration in wilderness, keep on gently guiding them when they seem to be lost, and lift them up when they seem to be in a hole but instead of trying to climb out of the hole, seem to be trying to dig it up further!

Forest from the Trees


My area of research eventually turned out to be the development of a unifying framework that casts all the previously considered separate approaches into a single formal structure, and turns each approach only a different perspective of looking at the same structure of information. Reaching to this higher level understanding was the essence of my contribution. But, the way one writes what one has done in PhD is not how one has discovered all the crucial elements. This challenge I have discussed in a separate post on how to make a more convincing thread of arguments (Conclusion vs Assumption in Research Writing).

This deep thinking that is required to connect disparate elements into a common higher level thread and identifying a unifying theme is my understanding of discerning the forest from the trees. Based on my hard work during those years, I think I learned the abstraction skills to connect disparate threads together and can identify relationships which are not only missed out by others. Even when I try to explain the connections they are unable to see them (may be the connections are too clear and concrete to me but too philosophical and abstract for others). It has often taken me years to first implement what I was saying so that the results come alive in front of the people, and only then they begin to appreciate what I had been saying for several years. I need to write a few examples of these in a separate post.

Multi-Disciplinary Focus [Epilogue]

 Browne’s research over five decades has spanned many domains in computer and computational science including many cross-disciplinary collaborations with physicists and engineers on topics ranging from binary black holes to control systems for prostate cancer treatments. 

Browne has attained fellow status in five different professional societies across several areas including the Association for Computing Machinery, the American Physical Society, the American Association for the Advancement of Science and the Institute for Constructive Capitalism.

JC's areas of focus spread far and wide. One day during our CODE Research group meetings the discussion shifted to the area of hardware performance from our usual area of focus of Visual and Parallel Programming. A group member who was a post doc (Dr Ravi Rao) made a remark saying that he thought none of the people in this group seems to be qualified to speak in this particular area. JC calmly listened to the member completing his remark and then said "Gentlemen, you may not know that around 2 decades ago in the early 1970s, this precisely was my area of research! Several of my PhD researchers's topic belonged to this particular area." Thereafter when we made a few inquiries, we found that his areas of specializations spread far and wide. Parallel and Visual Programming was only one of the groups. He had research groups in Aerospace Engineering, Electrical Engineering and even pure Mathematics and Physics running in parallel.  There were several areas which he had explored and left behind. Browne received the 2004 University of Texas at Austin Career Research Excellence Award for maintaining a superior research program across multiple fields over a 45-year career during which he supervised or co-supervised the Ph.D. research of 69 students in four different fields.

May be this was the inspiration for my writing the post: What does it Mean to be a PhD: Myths of specialization and departmental scope of expertise.

Now you may understand why my areas of focus as seen from the posts in my blog are spread far and wide spanning areas from IT, engineering, school education, learning psychology, politics, sociology, history, business etc. Interestingly enough, I may have been following my supervisor subconsciously. It was only yesterday, when I had finished the blog, I made this connection for the first time consciously, and then added the above part of this epilogue.

PS:

I was surprised to learn much later, that there were two JCs; JC that we graduate students knew and the JC that undergraduates knew. JC was helpful and considerate to the graduate students, however he was reputed to be unsympathetic and a hard task master for the undergraduates. Once I went to one of undergraduate class and saw why. It was in a big hall with several hundred students. There were several TAs assisting JC as he was flipping from that vast store of transparencies in his customary fast pace. Students did not have opportunity for much interaction except through the TAs; most of them from Far East whose accents often did not make much sense.  I think the differing in perception of graduate students vs undergraduates was mostly because of the large class sizes at undergraduate level and may be his primary interest being in conducting graduate research. It is interesting to note that even though UT Austin is a State University and often ranked among the top 10, the undergraduate class sizes were typically large (around 300-400) for such courses that most undergraduates are supposed to take. These courses were considered as weed-out courses where the university wanted to get rid of the weaker students by sophomore year that they were forced to admit under the state law.

Once he mentioned that he loathes the chairmanship of the department that is again coming back to him. The responsibility typically was rotated among the professors as they always preferred research and consultancy over admin responsibilities. He would now have to work for admin issues that he has no interest in. This represents an important area on which Pakistani universities need to work on: How to make consultancy and research more attractive than the administrative positions!

See Also: 



Profile of JC Browne :


James C. Browne is professor emeritus of computer science, research professor at ICES and chief technology officer for the Ranger system at the Texas Advanced Computing Center. He earned his Ph.D. in chemical physics from the The University of Texas at Austin.

Browne’s research over five decades has spanned many domains in computer and computational science including many cross-disciplinary collaborations with physicists and engineers on topics ranging from binary black holes to control systems for prostate cancer treatments. Browne’s current research interests span parallel programming and computation, performance optimization and fault/failure management for complex systems. One current project is enabling automation of performance optimization for multicore chips and multichip nodes of high performance computing. The tool implementing this automation, PerfExpert, has been adopted for use at several major high performance computing centers. Another is automation of fault and failure management for high performance computer systems.

Browne has attained fellow status in five different professional societies across several areas including the Association for Computing Machinery, the American Physical Society, the American Association for the Advancement of Science and the Institute for Constructive Capitalism. Browne received the 2004 University of Texas at Austin Career Research Excellence Award for maintaining a superior research program across multiple fields over a 45-year career during which he supervised or co-supervised the Ph.D. research of 69 students in four different fields.

7 comments:

  1. Excellent summary of your learning Dr Sahab. JazakAllah for sharing.

    ReplyDelete
    Replies
    1. Thanks for following and encouraging me to write.

      Delete
  2. This comment has been removed by a blog administrator.

    ReplyDelete
  3. This comment has been removed by the author.

    ReplyDelete
  4. This comment has been removed by the author.

    ReplyDelete
  5. Absolute genius! A good supervisor is a blessing for PhD scholar. Sir, your learning will be our guidelines in future Inshallah

    ReplyDelete
  6. You can hone your technical writing skills by looking at the writing and presentation styles of previous thesis which were previously published. Once you have finished the final draft for your thesis writing, reread it to fine tune your work. See more here mba dissertation examples

    ReplyDelete