The poet Langston Hughes wrote, in ``Theme for English B'' :
Go home and write \\
a page tonight. \\
And let that page come out of you--- \\
Then, it will be true. \\
\noindent He says it is just that simple, so here goes:\\
My name is Avani Wildani. I am a five foot tall mathy opera-loving vim-using
prolog-fangirl wheelchair-bound computer science graduate student who
would like to transfer into the UC Santa Cruz Computer Science Department.
And no, I'm not going to say that five times fast.
What do you know so far about me? I have decent grades, good GRE scores, and
a tendency to write embarrassingly long introductory sentences. I hope that
this description of my current research, my general research interests, and my
goals upon completing a Ph.D. tells you more about my enthusiasm for and
dedication to both research and learning.
I am currently a graduate student member of the Machine Learning group at the
University of New Mexico. I have chosen to apply to UC Santa Cruz to be near
friends and family who can care for me as I continue to study at a top-notch
research university. Though I am not about to leave my dreams of becoming a
research scientist, I am flexible on research areas and funding types if it
means I will be able to devote time to research instead of the trials of being
disabled and living alone.
My primary research interest is machine learning. I first became interested in
machine learning through a project I did for Sandia National Laboratories
during my senior year of college. I was the team leader for a group assigned
to write software that would take a multi-dimensional dataset cast into two
dimensions, cluster it, and analyze cluster validity. Sandia used our tool to
correlate data from infant leukemia studies, which made me feel good about my
work and drove me to seek more interdisciplinary projects.
In the UNM Machine Learning Group, I have worked on several projects. For
example, I have used Weka to model the correspondence between gene activation
and learned behaviors in mice. I have also implemented support vector
analysis with various kernels for very large data sets, and I used this to
evaluate relationships between fMRI regions of interest to predict
hierarchical relationships in functional brain areas for schizophrenic
I am currently working to identify schizophrenia subtypes using both fMRI data
and incomplete heterogeneous survey data collected from patients. Since it is
unlikely that this data is linear, I am modifying Ng et al.'s spectral
clustering algorithm to handle hidden values based on Sanguinetti's technique
for cross-entropy minimization for hidden variable analysis in Kernel PCA. I
hope to tie up the loose ends of this project before leaving UNM, and very
much hope that I can apply the techniques I have learned at UNM in the UC
Santa Cruz Machine Learning group. In particular, Professor Warmuth's work on
randomized PCA would be interesting to extend using kernel methods. Also, I
find Professor Haussler's work in bioinformatics and genomics interesting because
I prefer working with data that could give real world insight. I would enjoy
applying machine learning models to this area. Also, my recent research
called me to read Professor Haussler's paper on probabilistic PCA, and I'm
actively working on the analogous probabilistic interpretation for spectral
My ideal situation would be to work on bioinformatics and learning. Two
aspects of machine learning interest me: the thrill of watching screens of
data turn into meaningful patterns, and the possibility of one day gaining
insight into animal thought through building advanced artificial pattern
recognition systems. So you don't think that I am a crank, I should qualify
that I know how far-fetched this goal is. However, if distributed computing
and processor technology continue to grow as they have, I feel that one day
we'll be able to model a sufficiently complex system to gain some
understanding of ourselves.
In addition to my machine learning background, I have a working knowledge of
and interest in networks, particularly routing algorithms in non-traditional
network topologies, ubiquitous computing, and adaptive intrusion detection
systems. I am early enough in my graduate career that I would be willing (eager,
even) to test the waters outside of machine learning if I saw an interesting
problem come my way. In particular, I would love to know more about the
Hybrid Systems Modeling Framework for Data Communication Networks project (it
is not accessible at the moment).
At Harvey Mudd, I worked on network algorithms research with Professor Ran
Libeskind-Hadas. In particular, we studied graph algorithms and combinatorial
techniques to determine, as a function of the number of wavelengths and nodes
in our model, the minimum number of conversions to and from the optical
signals in ring-based network topologies. I also wrote a survey paper that
covered the state of split-packet, aka ``wormhole'' routing circa 2002.
At UNM, I was briefly involved in an scalable systems project that designed
boards to attach to porcupines to use 802.11b/g to monitor the animals even
when they burrowed. I was also a member of a networks research group at
Harvey Mudd headed by Professor Mike Erlinger. Our group finished implementations
of IDXP and BEEP in C and Java to create a working IDS that used a prototype
of IDMEF and IDXP. Professor Erlinger is a primary author of the IDMEF draft, and
I feel fortunate to have had the opportunity to help refine a protocol that is
going to grow up to be a RFC.
My plan after graduation is to go into academia or, if I can not get a
position in California, pursue a research position in government or industry.
I want to end up somewhere where I can continue to work on interesting
problems and have access to intelligent colleagues. Ideally, these colleagues
will be from a range of disciplines, since machine learning is nothing without
vast amounts of data from the experimental sciences, and I am much more
compelled by research projects that have the potential to help others.
I prefer academia because I enjoy both teaching and learning from my students.
As an undergraduate, I was responsible for grading and tutoring classes
ranging from upper-division Algorithms to Differential Equations. I graded
Discrete math for one of your current students (Dr. Greg Levin). I was also a
member of the Computer Science Staff, which is a group of student systems
administrators who maintained the SGI and Solaris boxes and helped students
and faculty work more efficiently. At UNM, I was the lead TA for a 200
student introductory Java class. I was in charge of leading labs, writing and
grading assignments, and supervising 4 other TAs. While I would prefer to get
back into research as soon as possible, I would not mind additional teaching
% I don't like this part!
Harvey Mudd College is a rigorous undergraduate institution that does not
inflate grades. I may have a 3.7 from UNM, but I am much more proud of my
2.8 from Harvey Mudd. This is a perfectly reasonable Mudd GPA that put me
almost squarely in the center of my graduating class of high school
valedictorians and National Merit finalists, and, additionally, I have
attached a list of the setbacks I overcame to graduate in four years.
I like the computer science department at UC Santa Cruz, I like the work you
are doing, I feel like my interests fit within your department, and I would
be glad to be a part of it. I guess it really is just that simple.\\ \\
\noindent With respect and thanks,
Update: Version 4.01, with many thanks to everyone who's helped on this! I just gave it to my advisor to rip apart... but I'm still really unhappy with the ending :-(
Thanks a lot for any and all comments. I'm going to back to being a stress-monkey now.