46.txt

This complete matlab for neural network

you get so you can design and implement a custom TMS algorithm (say) in an aft
ernoon. (Then you'll be debugging it on and off for the next six weeks, but th
at's how it is.) Sometimes making a standard package work can turn into a thes
is in itself.

Like papers, programs can be over-polished. Rewriting code till it's perfect, 
making everything maximally abstract, writing macros and libraries, and playin
g with operating system internals has sucked many people out their theses and 
out of the field. (On the other hand, maybe that's what you really wanted to b
e doing for a living anyway.)

9. Advisors
At MIT there are two kinds of advisors, academic advisors and thesis advisors.


Academic advisors are simple so we'll dispose of them first. Every graduate st
udent is assigned a faculty member as academic advisor, generally in his or he
r area, though it depends on current advisor loads. The function of the academ
ic advisor is to represent the department to you: to tell you what the officia
l requirements are, to get on your case if you are late satisfying them, and t
o OK your class schedule. If all goes well, you only have to see your academic
 advisor in that capacity twice a year on registration day. On the other hand,
 if you are having difficulties, your academic advisor may be able to act as a
dvocate for you, either in representing you to the department or in providing 
pointers to sources of assistance.

The thesis advisor is the person who supervises your research . Your choice of
 thesis advisor is the most important decision you'll make as a graduate stude
nt, more important than that of thesis topic area. To a significant extent, AI
 is learned by apprenticeship. There is a lot of informal knowledge both of te
chnical aspects of the field and of the research process that is not published
 anywhere.

Many AI faculty members are quite eccentric people. The grad students likewise
. The advisor-advisee relationship is necessarily personal, and your personali
ty quirks and your advisor's must fit well enough that you can get work done t
ogether.

Different advisors have very different styles. Here are some parameters to con
sider.

·       How much direction do you want? Some advisors will hand you a well-de
fined thesis-sized problem, explain an approach, and tell you to get to work o
n it. If you get stuck, they'll tell you how to proceed. Other advisors are ha
nds-off; they may give you no help in choosing a topic at all, but can be extr
emely useful to bounce ideas off of once you find one. You need to think about
 whether you work better independently or with structure. 

·       How much contact do you want? Some advisors will meet with you weekly
 for a report on your progress. They may suggest papers to read and give you e
xercises and practice projects to work. Others you may not talk to more than t
wice a term. 

·       How much pressure do you want? Some advisors will exert more than oth
ers. 

·       How much emotional support do you want? Some can give more than other
s. 

·       How seriously do you want to take your advisor? Most advisors will su
ggest thesis topics fairly regularly. Some can be depended on to produce sugge
stions that, if carried out diligently, will almost certainly produce an accep
table, if perhaps not very exciting thesis. Others throw out dozens of off-the
-wall ideas, most of which will go nowhere, but one in ten of which, if pursue
d with vision, can result in ground-breaking work. If you choose such an advis
or, you have to act as the filter. 

·       What kind of research group does the advisor provide? Some professors
 create an environment in which all their students work together a lot, even i
f they are not all working on the same project. Many professors get together w
ith their all their students for weekly or biweekly meetings. Will that be use
ful to you? Are the advisor's students people you get along with? Some student
s find that they construct important working relationships with students from 
other research groups instead. 

·       Do you want to be working on a part of a larger project? Some profess
ors divide up a big system to be built into pieces and assign pieces to indivi
dual students. That gives you a group of people that you can talk to about the
 problem as a whole. 

·       Do you want cosupervision? Some thesis projects integrate several are
as of AI, and you may want to form strong working relationships with two or mo
re professors. Officially, you'll have just one thesis supervisor, but that do
esn't have to reflect reality. 

·       Is the advisor willing to supervise a thesis on a topic outside his m
ain area of research ? Whether or not you can work with him or her may be more
 important to both of you than what you are working on. Robotics faculty at MI
T have supervised theses on qualitative physics and cognitive modeling; facult
y in reasoning have supervised vision theses. But some faculty members are onl
y willing to supervise theses on their own area of interest. This is often tru
e of junior faculty members who are trying to build tenure cases; your work co
unts toward that. 

·       Will the advisor fight the system for you? Some advisors can keep the
 department and other hostile entities off your back. The system works against
 certain sorts of students (notably women and eccentrics), so this can be very
 important. 

·       Is the advisor willing and able to promote your work at conferences a
nd the like? This is part of his or her job, and can make a big difference for
 your career. 

The range of these parameters varies from school to school. MIT in general giv
es its students a lot more freedom than most schools can afford to.

Finding a thesis advisor is one of the most important priorities of your first
 year as a graduate student. You should have one by the end of the first year,
 or early in the second year at the latest. Here are some heuristics on how to
 proceed:

·       Read the Lab's research summary. It gives a page or so description of
 what each of the faculty and many of the graduate students are up to. 

·       Read recent papers of any faculty member whose work seems at all inte
resting. 

·       Talk to as many faculty members as you can during your first semester
. Try to get a feel for what they are like, what they are interested in, and w
hat their research and supervision styles are like. 

·       Talk to grad students of prospective advisors and ask what working fo
r him or her is like. Make sure you talk to more than one student who works wi
th a particular advisor as each advisor has a large spectrum of working styles
 and levels of success in interaction with his or her students. You could be m
isled either way by a single data point. Talk to his or her first year advisee
s and his seventh year advisees too. 

·       Most or all faculty member's research group meetings are open to new 
grad students, and they are a very good way of getting an idea of what working
 with them is like. 

AI is unusual as a discipline in that much of the useful work is done by gradu
ate students, not people with doctorates, who are often too busy being manager
s. This has a couple of consequences. One is that the fame of a faculty member
, and consequently his tenure case, depends to a significant extent on the suc
cess of his students. This means that professors are highly motivated to get g
ood students to work for them, and to provide useful direction and support to 
them. Another consequence is that, since to a large degree students' thesis di
rections are shaped by their advisors, the direction and growth of the field a
s a whole depends a great deal on what advisors graduate students pick.

After you've picked and advisor and decided what you want from him or her, mak
e sure he or she knows. You advisor may hear ``I'd like to work with you'' as 
``Please give me a narrowly specified project to do,'' or ``I've got stuff I'd
 like to do and I want you to sign it when I'm done,'' or something else. Don'
t let bad communication get you into a position of wasting a year either spinn
ing your wheels when you wanted close direction or laboring under a topic that
 isn't the thing you had your heart set on.

Don't be fully dependent on your advisor for advice, wisdom, comments, and con
nections. Build your own network. You can probably find several people with di
fferent things to offer you, whether they're your official advisor or not. It'
s important to get a variety of people who will regularly review your work, be
cause it's very easy to mislead yourself (and often your advisor as well) into
 thinking you are making progress when you are not, and so zoom off into outer
 space. The network can include graduate students and faculty at your own lab 
at others.

It is possible that you will encounter racist, sexist, heterosexist, or other 
harrassment in your relationships with other students, faculty members, or, mo
st problematically, your advisor. If you do, get help. MIT 's ODSA publishes a
 brochure called ``STOP Harrassment'' with advice and resources. The Computer 
Science Women's Report, available from the LCS document room, is also relevant
.

Some students in the lab are only nominally supervised by a thesis advisor. Th
is can work out well for people who are independent self-starters. It has the 
advantage that you have only your own neuroses to deal with, not your advisor'
s as well. But it's probably not a good idea to go this route until you've com
pleted at least one supervised piece of work, and unless you are sure you can 
do without an advisor and have a solid support network.

10. The thesis
Your thesis, or theses, will occupy most of your time during most of your care
er as a graduate student. The bulk of that time will be devoted to research , 
or even to choosing a topic, rather than to the actual writing.

The Master's thesis is designed as practice for the PhD thesis. PhD-level rese
arch is too hard to embark on without preparation. The essential requirement o
f a Master's thesis is that it literally demonstrate mastery: that you have fu
lly understood the state of the art in your subfield and that you are capable 
of operating at that level. It is not a requirement that you extend the state 
of the art, nor that the Master's thesis be publishable. There is a substantia
l machismo about theses in our lab, however, so that many Master's theses do i
n fact contribute significantly to the field, and perhaps half are published. 
This is not necessarily a good thing. Many of us burn out on our Master's work
, so that it is notorious that MIT Master's theses are often better than the P
hD theses. This defeats the preparatory intent of the Master's. The other fact
or is that doing research that contributes to the field takes at least two yea
rs, and that makes the graduate student career take too damn long. You may not
 feel in a hurry now, but after you've been around the Lab for seven years you
'll want out badly. The mean time from entrance to finishing the Master's is t
wo and a half years. However, the CS department is strongly encouraging studen
ts to reduce this period. If a Master's topic turns out to be a blockbuster, i
t can be split into parts, one for the Master's and one for a PhD.

To get some idea of what constitutes a Master's thesis-sized piece of research
 , read several recent ones. Keep in mind that the ones that are easy to get a
t are the ones that were published or made into tech reports because someone t
hought they extended the state of the art---in other words, because they did m
ore than a Master's thesis needs to. Try also reading some theses that were ac
cepted but not published. All accepted theses can be found in one of the MIT l
ibraries. PhD theses are required to extend the state of the art. PhD thesis r
esearch should be of publishable quality. MIT machismo operates again, so that
 many PhD theses form the definitive work on a subarea for several years. It i
s not uncommon for a thesis to define a new subarea, or to state a new problem
 and solve it. None of this is necessary, however.

In general, it takes about two to three years to do a PhD thesis. Many people 
take a year or two to recover from the Master's and to find a PhD topic. It's 
good to use this period to do something different, like being a TA or getting 
a thorough grounding in a non-AI field or starting a rock and roll band. The a
ctual writing of the PhD thesis generally takes about a year, and an oft-confi
rmed rule of thumb is that it will drag on for a year after you are utterly si
ck of it.

Choosing a topic is one of the most difficult and important parts of thesis wo
rk.

·       A good thesis topic will simultaneously express a personal vision and
 participate in a conversati