Nonlinear data analysis encompasses a broad range of
techniques that are based in numerical optimization. This course
will cover some techniques that specifically address sets of data
in vector spaces.
To test, implement, and analyze this material, we will use MATLAB
as an interactive tool and programming language. Students are
expected to learn basic MATLAB on their own.
For basic material in data analytics, students can expect to be
elements of optimization for scalar arguments
basic methods for optimization in Euclidean spaces
applications in "shallow" neural networks
basic methods for constrained optimization using KKT conditions
applications in constrained regression and SVM classification
The course contents, including the schedule, notes, and
supplementary material, are provided in
the onQ learning management
Methods for computational optimization, particularly examining
nonlinear functions of vectors. Topics may include: unconstrained
optimization; first-order methods; second-order methods; convex
problems; equality constraints; inequality constraints;
applications in machine learning.
Each student is expected to have read the relevant class notes and
viewed the recorded lecture material before attending class, and
to have understood the important definitions and concepts. The
class time will concentrate mainly on the use of concepts to solve
problems as exercises. Definitional material will not be covered
because students are expected to learn the definitions before
Each student is responsible for transcribing class material. The
instructor will not scan or otherwise copy the lecture notes,
which are intended to reflect and supplement the on-line class
notes available in onQ.
Recent Course News (as of April 29, 2018)
The course will next be offered in September, 2020