Ph.D. Interviews in 2015
"Coding and Machine Learning for Good Understandings"
Muhammad Awais SHAFIQUEiD3)
£ Talking with Junji at Lab.
Eiji: Let me start by asking, Why do you choose the destination to study
at the University of Tokyo?
Shafique : There are a lot of factors involved. Firstly, I was greatly
motivated by my Professor (graduated from Todai) at my home university.
Secondly, at the time of my application, the department of Civil
Engineering at Todai was ranked 3rd in the world. And thirdly, I had a
lot of friends already studying at Todai so it was much more attractive
than the other universities.
Eiji: We have interests in continuing difficulties of many Asian countries
and can gear to the realities in transportation and urban planning in
Pakistan. How did you pick the topic of your thesis? It is most important
thing to make and review a research proposal.
Shafique : My original research proposal was on the application of
Transportation Demand Management (TDM).
Eiji: I can look back upon ur first research proposal. The comprehensive
explanation of the proposed research was addressed not to laymen but to
other specialists in your field. TDM is not bad proposal but too general
as research proposals.
Shafique: After Prof. Hato was finalized as my supervisor, I had a
number of meetings with him. In the meantime, I studied the research
done by Prof. Hatofs lab. For that purpose, the students of that lab
helped me a lot as most of the research was documented in Japanese
language. After completing the necessary background study and having
extensive discussions with Prof. Hato, I ultimately decided to work on
travel mode detection using sensors.
Eiji: It is difficult for PhD students first without a guidance of researcher
who have experience in research. BinN studies unit has already written
some papers concerned to data sensing methodologies and probe person
technologies for Transportation Research Part C. Moreover Imaizumi has
tried Ada-Boost algorithm for mode detection and discussed initial
research approach with you animatedly. Explain about your research
idea in detail.
Shafique: The idea was simple yet quite interesting. To detect the mode
of transportation used by a participant by analyzing the multiple sensorf
s data collected from his/her smartphone or any other wearable device.
Although, this is not a novel idea but is relatively very recent,
therefore requires a lot of research. The aim was to improve the
detection accuracy of the methodology, while decreasing the battery
Eiji: A wearable, behavioral context information-measuring instrument,
for re-estimating label information such as transport mode from multiple
sensors based on learning models. The numerical values observed by these
sensors differed greatly among locations or transportation modes, revealing
the high possibility of automatic identification of transportation modes
using machine leaning which are naive bayse, random forest, SVM, adaboost
and decision tree. Moreover you have a good view, because the Probe Person
survey has a tradeoff between survey accuracy and efficiency of battery.
Shafique: To achieve these goals, machine learning algorithms were
decided to be used. It was also decided to experiment with Multinomial
Eiji: Incorporating discrete choice models into machine learning algorithm
is unique idea to solve problems of survey accuracy. Important idea is
features from two methods are mutually orthogonal. Did you join the
conference and submit to the JournalH And is it difficult?
Shafique : Yes, I have submitted a number of papers both to conferences
and journals. The difficulty level increases with increase in the impact
factor of journal or prestige of conference. I have presented my papers
at conferences in Japan as well as in other countries including France,
Spain, Greece, USA and the Netherlands. Among them, the most prestigious
were CUPUM 2015 organized by MIT, USA and EWGT 2015 organized by TU
Delft, the Netherlands. Coming to the journals, I have
published/submitted in top journals in the field of Transportation
Engineering like Transportation Research Part C, Transportation etc.
Publishing in journals is a very difficult task, requiring a lot of
patience as the time for review and process may range from 6 months to a
year or even more than that.
£ Taking summer course of Behavior Modelling, with BinN's Friends.
Eiji: PhD students should respond positively to join the conference and
publish, you give a positive impression. And you have many oppotunities
to discuss your research topics with Kay Axhausen holding Transportation
chief editor and a lot of researchers in BinN international reserach
seminar. All of research is not entirely smooth. How do you battle with
difficulty in your research, make a code and study theories of machine
Shafique : Of course there are times when one gets stuck at something
and ultimately gets frustrated. I too had my due share. Usually, when I
got frustrated, I would turn off my computer and head back home to spend
time with my wife. It is that when I am relaxed, ideas pour in. So the
next day I would again tackle the problem with high spirits and new
ideas. Furthermore, I took regular breaks during my research and would
visit a friend or invite a friend for a cup of coffee. These informal
meetings helped me to release the tension and made me relaxed. When I
started my research, I was not good at coding. In fact, it had been
years since I wrote my last code. Therefore, obviously I was in
difficulty, but I overcame it with will and hard work.
Eiji: Our research is made mostly of coding. The remaining 10% of the
code accounts for the other 90% of the development time. In our lab,
we provide summer course of travel behavior modelling and R for young
Shafique : I read literature about gRh and gJavah, and went through
the previous codes written by other students. It all helped me to develop
my understanding. Similar was the case with Machine Learning algorithms.
It was the first time that I was reading about them. Extensive study
of material present online assisted me to learn this new knowledge,
which was later used in my research.
Eiji: How about thesis defense? You looks getting too nervous when your
Shafique : My supervisor, co-supervisors and lab. Professors had all
provided me with so much feedback that I was fairly relaxed after the
completion of my final presentation and was eagerly waiting for my turn.
Preparation gives you confidence, so I was pretty much confident and
this helped me to deliver a good presentation.
Eiji : You strived harder for some goal, your family and country.
My last question is about Japanese student and UT and Japan.
Shafique : I have found Japanese students to be very helpful, polite and
hardworking. They were always there to assist me whenever I had some
difficulties either in the university or outside. Regarding Tokyo, I
consider these 3 years as one of the best times in my life. Japan is a
beautiful country and the Japanese people are all very nice. I would
also like to mention that wherever I went, the Japanese people were
always kind to me and always accommodated my special needs as a Muslim,
especially my Professors and other students always gave due
consideration to the fact that I had some dietary restrictions.
TNX 6, arigato sensei!
£ at the international conference in Delft.