Short Courses
In addition to the student talks and plenary sessions, we are offering a series
of short courses in conjunction with the conference. These courses will be offered
during the afternoon of Friday, April 24, 2015 at 1:00pm. The short courses are
open to all registrants, and there is no need to pre-register for any particular
coruse. Upon arrival at the conference, you may choose in which course, if any, you
would like to participate. The title and descriptions of each course are given
below.
Understanding Data: From Math (Probability) to Statistics to Machine Learning,
and Back to Math (Topology)
Chris Overton, Ayasdi (Advanced Analytics & Big Data Analytics)
In this session, we will survey a variety of techniques for modeling, forecasting, and
discovering patterns in uncertain data. We will present and critique examples using
probabilistic models, regression, hierarchical clustering (along with map-reduce), and
finally the topological mapper algorithm. For each of these techniques, we discuss a high-level
theoretical overview, practical implementation and benefits, as well as demonstrations of what
they fail to detect. Data science is now recognized as an important subject, but it is evolving
rapidly, and people use the term in very different ways. By providing historical and interdisciplinary
background, we hope to help students plan better for how to develop expertise in this challenging area.
In particular, needed CS skills are also evolving rapidly. Though this is a small part of the talk, we
survey current tools (flat-file etl, sql db's, data frames, Hadoop, Spark) and suggest ways to avoid
getting trapped in “data munging” so that one is able to proceed to data science. We close by discussing
and showing visualizations of the mapper algorithm, which is used by Ayasdi to help customers in several
industry verticals make new sense of their “big data.”
The Catalan Numbers
Dr. Ralph Grimaldi, Rose-Hulman Institute of Technology
One of the most ubiquitous number sequences in mathematics, the Catalan numbers arise in
problems dealing with lattice paths, trees, balancing parentheses, compositions, data structures,
pattern avoidance, graph theory, coin arrangements, and other areas. In this minicourse we shall
investigate some of these applications and also look at some of the properties satisfied by these
numbers.
Detection Theory on Random Graphs
Dr. Thomas Mifflin, Metron Scientific Solutions
Slides Available Here
After the attacks on September 11th, 2001, military and intelligence communities used link analysis
to attempt to localize key individuals that masterminded the attack. Link analysis represents relationships
among individuals as a graph. Immediately before and after the attack our adversaries took great care into
blending into the background of the societies where they resided. The technical question became can we
distinguish between the "noise" of relationships in normal societies from the "signal" of the structure
of terrorist organizations? This mini-course will show how classical detection theory based on the
Neyman-Pearson lemma from statistics (likelihood ratio) can be applied to models of signal and noise in
certain classes of random graphs. The course concentrates on the simplest model of random graphs, the
Erdos-Renyi model, G(n,p), where n is the number of vertices in the graph and p is the probability that any
two vertices are linked with an edge. The course ends with showing selected results from other random graph
models, such as the random bipartite projection (Kevin Bacon) model.
|