EE5585 Data Compression (Spring 2015)
Instructor: Arya Mazumdar Email:
arya@umn.edu
8:15 A.M. - 9:30 A.M. We,Th AmundH 240
Instructor availability: By
appointment and after class.
Blurb: This course introduces students to data
compression algorithms and the scenarios to which they are applied. Some of the
popular data-compression techniques that will be covered includes Huffman
Codes, Lempel-Ziv Universal coding, BH coding and Transform coding (JPEG
etc.). The course will begin with an introduction to the fundamental problems
of data compression and their mathematical formulations. This will be followed
by a study of the methods that underlie most data compression algorithms, such
as information theory and quantization. We will cover topics that are relevant
to today's Big Data framework - and the course will involve projects on cutting
edge topics - compressed sensing, complex data- analysis and clustering.
Grading Policy: Homeworks 20%, Midterm 30%, Project 40%, and Presentation
10%. No Final Exam. Project result publishable in conference/journal earns
automatic A.
Syllabus:
Subject
to updates.
¤ Compressible data, entropy, information theory, variable
length coding
¤ Universal coding: Lempel Ziv,
Burrows-Wheeler
¤ Complexity: Turing Machine and Kolmogorov Complexity,
Complexity of a file
¤ Quantization and Rate Distortion: Scalar, Vector, LBG,
Lloyd, K- Means
¤ Compressed Sensing - Sparse Recovery - Basis Pursuit
¤ Distributed compression - compression in Big Data - query
efficiency, update-efficiency
¤ Transform coding, Wavelets and JPEG
Course Projects: Each team should contain 1-3 students. The Students will
choose a topic on their own or will select one from the following project
topics (subjects to updates).
¤ Big Data
Compression
¤ Compression of
Biological Data and DNA sequences
¤ Data
Compression and Machine Learning
¤ Compressive
Sampling
¤ Distributed
Data Compression/ Networked Data Compression
Homeworks:
Lectures:
Handouts: JPEG,
Burrows-Wheeler