Potential Grad students:
Please excuse the "canned" nature of this advice. I receive several
hundred inquiries such as yours every year, and this turns out to
be the most convenient way to respond.
[Cribbed from Lera Boroditsky's page!]
Should I email you to let you know I am applying?
No. Really. This advice is specific to me (other professors may want
you to get in touch). The only way I consider applications is through
our organized admissions process. I won’t be able to look at any
materials that you send me directly. So please, put that good proactive
energy into your official application. I take admissions very
seriously and I will read your official application very closely once I
receive it through our admissions process. There is no reason to
email me before you apply and no advantage to doing so. Unfortunately,
I typically do not have time to read or reply to individual emails
Can I come visit your lab/call you to discuss my application?
In the interest of fairness, I ask everyone to just apply and go
through the same application process. I do not set up special meetings
or phone calls with folks before they apply because it would not be
fair to all of the other people who apply who don't get a special
meeting or phone-call. So, please just apply. If you are selected by
the admissions committee, you will be invited to our departmental open
house, where you'll get to meet all the folks in the department and
spend time with the graduate students and learn a great deal about what
we all do here and what life here is like. So, please apply, and if the
admissions committee selects you, you will be invited. Unfortunately my
lab can't accommodate visits otherwise.
[END Crib from Lera's page!]
First, I can't accept new students on an individual basis - you have to
go through the regular admissions process. Information on this and our
department can be obtained from:
Please read the requirements first!
We turn away a large number of highly qualified people each year, so if
you don't get in, it isn't necessarily your fault! Our strongest
competitors are schools like Stanford, Berkeley, U. Washington, CMU and
MIT. So, again, our students are the cream of the crop.
If you are interested in what I do, that is, Cognitive Science, you
should also consider the Cognitive
Science department at UCSD which has quite a few computational
people, Virginia de Sa (Neurocomputational models of learning and
vision, Brain Computer Interfaces), Jeff Elman (Neural networks models
of Language and Development), Jim Hollan (HCI), Ed Hutchins (HCI,
distributed cognition), David Kirsh (HCI and AI), and Angela Yu
(computational cognitive neuroscience).
A third alternative is the Interdisciplinary
Ph.D. Program in Cognitive Science, of which I am currently the
Director. You can enter this program from any of eight different
departments on campus, including ours and Philosophy.
By the way, here are a few good sites for "going to grad school" advice:
illustrated guide to a Ph.D.
Phil Agre's "Advice for
Undergraduates Considering Graduate School.
Often cited as "the best" advice, is Maries desJardins' article "How to succeed in
Graduate School: A guide for students and advisors"
Here is David Chapman's "How to do
research at the MIT AI Lab," which is more widely applicable than
its title would suggest:
And here is a short list of concise
advice by Wanda Pratt.
Most of these have further links. Please let me know if any links are
I have included here a list of other schools, along with my research
Starting with UCSD:
I am the main person doing Cognitive Science in my department. There
are people here doing machine learning and cognitive science in
my department (CSE), Electrical and Computer Engineering (ECE),
Cognitive Science, Biology, Philosophy, Neuroscience and Medicine.
Also, Terry Sejnowski is here (he is in most departments on campus, at
least it seems that way). There's probably others I have
forgotten to mention.
I do work in several areas, with the main goal being trying to
understand the neural basis for the phenomenon. The list is pretty
spread out. We have some nice results (see my
Another major interest is multi-layer Independent Components Analysis
(mainly due to my excellent former graduate student, Honghao Shan), and
novel approaches to ICA that are biologically plausible.
Another major interest is in recurrent nets. I have done work (or my
students have done work) in learning simple procedures (especially
arithmetic), face and emotion recognition, central pattern generators
(in particular, the lobster stomatogastric ganglion, but more recently,
the leech swim circuit), image compression, nonlinear dimensionality
reduction, active selection of training examples, an area that for me
has now morphed to "selective attention", information retrieval, speech
recognition with emphasis on adapting to speaking rate, word sense
Again, most of what I have done lately is face and object processing,
with an emphasis on two approaches: Neural nets, and Bayesian models.
Here's part of my research statement:
"My research is strongly interdisciplinary. It concerns using neural
networks and other formal approaches as computational models applied to
problems in Cognitive Science & Artificial Intelligence,
Engineering and Biology. The major thrust of my research over the
review period has been in cognitive modeling of perceptual phenomena,
as well as experimental approaches to testing those models in
humans. My ultimate goal is to understand how the mind works,
given that the only "hardware" it has to run on is our brains. I
am committed to the view that there is a rich and productive
interaction between computer science and related disciplines. On
the one hand, I believe that computer science has much to offer to the
understanding of cognitive and biological sciences. Computer
science offers not only methodological tools for studying complex
phenomena, but provides a theoretical basis for understanding
computation as it may appear in other domains. At the same time, I
believe that the study of "real world phenomena" offers a challenge to
computer science. Attempting to understand the computational
underpinnings of domains such as face and object recognition in humans
can stimulate the development of new theories and techniques which
enrich computer science."
Other CSE people:
David Kreigman is one of the most widely cited experts on the subject
of face recognition, a crucial component of vision-based security
systems for human-computer interaction as well as homeland security
purposes. In object recognition, David revolutionized the field with
his work on the geometry of curved surfaces to recognize objects from
their occluding contours, silhouettes and aspects.
Charles Elkan has done excellent work in cost sensitive learning,
clustering algorithms, finding sequences in biological databases, and
genetic algorithms applied to computational finance. He caused a great
deal of controversy with his paper on "The paradoxical success of fuzzy
Sanjoy Dasgupta has developed some of the first efficient, provably
correct algorithms for canonical statistical tasks (especially related
to clustering) on high dimensional data sets.
Serge Belongie is working on content-based image and video retrieval--a
new way to search and analyze the output from the growing number of
digital cameras everywhere (since the volume of images and video far
outstrips anyone's ability to log all the video).
Mohan Paturi has analyzed the basins of attraction in Hopfield
networks, threshold circuit complexity, and has been working on fast
New to our department in the last few years are Yoav Freund, who
co-discovered AdaBoost, one of the most widely used machine learning
algorithms in recent years, and Lawrence Saul, co-inventor of Local
Linear Embedding, which is a new and effective way to find the low
dimensional manifold that high dimensional data "lives in." Both of
these researchers have many other interests; I have just referred to
what they are most famous for.
Terry Sejnowski is here at the Salk Institute, and has a wide range of
interests, from modeling the hippocampus, to face recognition to speech
recognition to motion perception to schizophrenia.
I listed the Cog Sci folks above.
In ECE, Gert Lanckreit is a new addition in machine learning, a former
student of Mike Jordan's. Nuno Vasconcelos studies statistical
approaches to computer vision. Pamela Cosman works on image
processing and data compression, Alon Orlitsky works on
compression, machine learning, and speech recognition, Tony
Sebald on neural networks for control, Sing Lee, Clark Guest, and Shaya
Feinman work on optoelectronic implementations of neural nets,
Ken Kreutz-Delgado works on applications of neural nets to robotic
control, especially the inverse kinematics problem. Mohan Trivedi is a
world-famous expert on machine vision, intelligent systems, and
In Biology, Bill Kristan has done extensive work in modeling
invertebrate systems, especially "Behavioral Choice": how a simple
invertebrate system "decides" what to do next. We worked together on
leech swimming through our shared grad student, Adam Taylor.
In Psychology, there are a number of folks who use computational
methods to understand how the brain works: Don MacLeod, Ed Vul,
David Huber (leaving soon for Kent in England), Adam Aaron, and John
People in the Med school have used neural nets for diagnosing eye
problems. Larry Squire works on memory and has done some modeling of
Other schools (quite out of date!):
Toronto: Geoff Hinton, Radford Neal, Rich Zemel, (who am I
MIT (Brain and Cognitive Sciences Dept.) has Josh Tenenbaum, Tommi
Jaakkola (I could have placed him in the wrong department). Also,
Tommy Poggio. Mike Jordan, while he was at MIT, claimed that
among him, Tommy Poggio and others, their students and postdocs, there
are 40-50 researchers at MIT working on learning, and so MIT is the
center of the machine learning Universe, just as it is the center of
the Universe for everything else.
U Mass Amherst (COINS): Andy Barto reinforcement learning (now retired).
University of Alberta: Rich Sutton reinforcement learning
CMU (Psych or CS Dept.): Dave Touretzky, Dean Pomerleau, Scott Fahlman
Dave Plaut, Mike Lewicki, Tom Mitchell, and many more.
U. of Rochester: Robbie Jacobs, Brain and Cognitive Sciences. James
Allen, Comp Sci.
Indiana U. (CS or Linguistics Depts.): Michael Gasser (language
production), Bob Port (speech recognition, recurrent networks), John
Krushke (human categorization) [this list is out of date - there are
U. of Colorado Boulder (CS Dept.): Mike Mozer and Randy O'Reilly.
Johns Hopkins: Paul Smolensky
Colorado School of Mines: Lori Pratt
Brandeis: Jordan Pollack (language, recurrent nets)
Columbia: Larry Abbott, Eero Simoncelli,
Stanford: Daphne Koller, Andrew Ng, and Bernard Widrow CS/EE.
Cal Tech: Carver Mead, Christof Koch (Computational Neural Systems)
USC: Stefan Schaal, Bart Mel, Michael Arbib, Ladan Shams,...
UCIrvine: (There are now a LOT more people than this list:) Pierre
Baldi, Eric Mjolsness, Padhraic Smyth, Mark Steyvers,...
UCLA: Adnan Darwiche, Mike Dyer, Stefano Soatto, Demetri Terzopoulos,...
UCSC (Psych): Alan Kawamoto (language), in CS, Dave Haussler (Learning
theory, bioinformatics applications), Manfred Warmuth.
Berkeley has Peter Bartlett, Jerry Feldman, David Forsyth, Marti
Hearst, Mike Jordan, Daniel Klein, Jitendra Malik, Stuart
Russell, Bob Wilensky. Leo Breiman in Statistics.
PS: A blurb from Sue Becker on McMaster U:
From: Sue Becker <firstname.lastname@example.org>
Subject: Openings for graduate students
To: connectionist <Connectionists@cs.cmu.edu>
Students interested in pursuing a graduate degree in psychology with
an emphasis on connectionist modelling are encouraged to apply to
McMaster University. The Department of Psychology was recently
appraised by an external review committee as being among the very
top experimental psychology departments in Canada. A number of the
faculty are conducting research into the neurobiological bases
of perception and cognition, and share a growing interest in
employing connectionist/computational models in their work. There
is ample opportunity for students to collaborate with two or more
members of this group, using a combination of experimental and
To apply, contact Ms. Erie Long, Graduate Secretary, Department of
Psychology, McMaster University, 1280 Main St. West, Hamilton, Ont.
Canada, L8S 4K1
tel: (905)525-9140 ext 23000
Additionally, feel free to contact any of the faculty listed below
for further information about the department or to discuss the
possibility of thesis supervision:
Dr. Sue Becker, email@example.com
Research on unsupervised learning in neural networks,
connectionist models of perceptual development and of
and cortical learning mechanisms.
Dr. Kathryn Murphy, firstname.lastname@example.org
Research on the development, plasticity, and organization
visual cortex and central visual pathways -- using
ranging from cellular neurobiology to computational
Dr. Ron Racine, email@example.com
Investigations into the synaptic mechanisms of
long-term potentiation model of memory is used to study the
mechanisms by which patterned activation can induce
in connection weights. Various forms of the Hebb
rule appear to
apply to these potentiation phenomena. The
experimental data are
used to guide modelling efforts, particular with respect to
associative learning involving the neocortex.
Dr. Larry Roberts, firstname.lastname@example.org
Research on associative learning, with a special interest
cortical mechanisms assessed by high resolution
(EEG) and neuromagnetic (MEG) recording in human
interest includes modelling of learning systems and
of neural networks to problems of signal processing.
Dr. Judith Shedden, email@example.com
Research on human attention and spatial processing, using
behavioural measures and the brain-imaging techniques of
resonance imaging and event-related potentials.
Faculty outside of the psychology department whose research
involves computational modelling:
Dr. Simon Haykin, firstname.lastname@example.org (Department of
Electrical and Computer Engineering)
Modelling of the communication process between sites in the
nervous system, which would include the effects of
nonstationarity. This work would involve time frequency
analysis, and neural networks.
Dr. David Jones, email@example.com (Computer Science Department)
Research on computational models of biological vision
including models of early cortical visual processing and