Network Analysis and Modeling

CSCI 5352, Fall 2019
Time: Monday, Wednesday, Friday, 1:00pm - 1:50pm
Room: ECCR 105

Lecturer: Dan Larremore
Office: ECCE 1B06
Office hours: Mondays, 2:00 - 4:00 PM
Email: daniel.larremore

Teaching Assistant: Hunter Wapman
Office: ECAE 128, Table 2
Office hours: Wednesdays, 10:30 - 12:00 and 3:00 - 4:30

Syllabus

Description
Course work and grading
Schedule and lecture notes
Problem sets
Supplemental readings

Description
Network science is a thriving and increasingly important cross-disciplinary domain that focuses on the representation, analysis, and modeling of complex social, biological and technological systems as networks or graphs. Modern data sets often include some kind of network. Nodes can have locations, directions, memory, demographic characteristics, content, and preferences. Edges can have lengths, directions, capacities, costs, durations, and types. And, these variables and the network structure itself can vary, with edges and nodes appearing, disappearing and changing their characteristics over time. Capturing, modeling and understanding networks and rich data requires understanding both the mathematics of networks and the computational tools for identifying and explaining the patterns they contain.

This graduate-level course will examine modern techniques for analyzing and modeling the structure and dynamics of complex networks. The focus will be on statistical algorithms and methods, and both lectures and assignments will emphasize model interpretability and understanding the processes that generate real data. Applications will be drawn from computational biology and computational social science. No biological or social science training is required. (Note: this is not a scientific computing course, but there will be plenty of computing for science.)

Prerequisites
Recommended: CSCI 3104 (undergraduate algorithms) and either CSCI 3022 or APPM 3570 (key knowledge: probability), or equivalent preparation.

Note: An adequate mathematical and programming background is mandatory. The concepts and techniques covered in this course depend heavily on basic statistics (distributions, Monte Carlo techniques), scientific programming, and calculus (integration and differentiation). Students without sufficient preparation will struggle to keep up with the lectures and assignments. Students without proper preparation may audit the course.

Text
Required (available at the CU Bookstore):
1. Networks: 2nd Edition by M.E.J. Newman
2. Pattern Recognition and Machine Learning by C.M. Bishop.

Optional:
1. All of Statistics by L. Wasserman
2. Numerical Recipes
3. Networks, Crowds and Markets by D. Easley and J. Kleinberg
4. Error and the Growth of Experimental Knowledge by D.G. Mayo.

Course work and grading
Attendance to the lectures is required.

Most of the class will be standard graduate-style lectures by the instructor. These will be supplemented by guest lectures on special or advanced topics, and class discussions of selected papers drawn from the networks literature. Problem sets will develop and extend topics presented in class, and will introduce additional topics not covered in class. Performance on the problem sets will be the major component of evaluation. There are no written examinations in the course, and thus students are expected to spend serious quality time on the problem sets. Additional details are given in the syllabus.

Problem sets: There will be 6 problem sets. Each will include some mathematical and some computational problems. Problem sets will be due roughly every two weeks. Programming components of the problem sets may be completed in any reasonable imperative language, and students are not expected to code everything from scratch (using available network libraries is okay).

See the syllabus for more details about formatting and submitting your solutions, for advice about how to get maximum points, and for the class policies on collaboration and on late submissions. Students that are unsure about whether something is permitted under the policies described in the syllabus should consult with the instructor well before a particular deadline.

Class project: The purpose of the class project is to formulate and explore a research question of the student's devising related to network analysis and modeling. Students may work in small teams. The deliverables are (i) a short (10 minute) in-class presentation of the project results, and (ii) a 10-page writeup. See the syllabus for more details.

Grading: See the syllabus.

Tentative Schedule
Week 1 : Introduction and overview (Lecture 0 and Lecture 1)
Week 2 : Measures of structural importance (Lecture 2)
Week 3 : Random graphs I: homogeneous degrees (Lecture 3, webweb, G(n,p)+webweb, PercoVIS.)
Week 4 : Random graphs II: heterogeneous degrees (Lecture 4)
Week 5 : Large-scale structure I: assortativity and modularity (Lecture 5)
Week 6 : Large-scale structure II: stochastic block models (Lecture 6, SBM+webweb, slides)
Week 7 : Spreading processes on networks (Lecture 7, Kleinberg Ch 24)
Week 8 : Ranking in directed networks and pairwise comparisons (Lecture 8)
Week 9 : Wrangling network data I: sampling (Lecture 9)
Week 10 : Wrangling network data II: data, statistics, and tests Lecture 10)
Week 11 : Spatial networks (Lecture 11)
Week 12 : Growing networks (Lecture 12)
Week 13 : Dynamic networks
--- Fall break ---
Weeks 14-15 : Project presentations and Wrap up

Problem Sets

Problem set 1 [data files linked via Piazza]

Problem set 2

Problem set 3

Problem set 4

Problem set 5

Problem set 6

Supplemental Readings

Week 1:

Week 2:

Week 3:

Week 4:

Week 5:

Week 6:

Week 7:

Week 8:
See links in Lecture 8 notes.

Week 9:

Week 9 Bonus:

Week 10:

Week 11:

Week 12:

Week 12:

Week 13:

Network Tools
NetworkX, network analysis package (Python)
igraph, network analysis tools (Python, C++, R)
graph-tool, network analysis and visualization software (Python, C++)
GraphLab, scalable network analysis (Python, C++)

Network Visualization
Cytoscape, network visualization software
yEd Graph Editor, network visualization software
Graphviz, network visualization software
Gephi, network visualization software
graph-tool, network analysis and visualization software
webweb, network visualization tool joining Matlab and d3
MuxViz, multilayer analysis and visualization platform

Network Data Sets
The Colorado Index of Complex Networks (ICON; more than 4000 graphs)
US Census Education-Employment network (social, bipartite, weighted)

Other Courses on Networks
Network Theory (University of Michigan)
Statistical Network Analysis (Purdue University)
Networks (Cornell University)
Networks (Harvard University)
Social and Economic Networks: Models and Analysis (Coursera / Stanford)
Social Network Analysis (Coursera / University of Michigan)
Social and Information Network Analysis (Stanford)
Graphs and Networks (Yale)
Spectral Graph Theory (Yale)
The Structure of Social Data (Stanford)

Resources
LaTeX (general) and TeXShop (Mac)
Matlab license for CU staff (includes student employees)
Mathematica license for CU students
NumPy/SciPy libraries for Python (similar to Matlab)
GNU Octave (similar to Matlab)
Wolfram Alpha (Web interface for simple integration and differentiation)
Introduction to the Modeling and Analysis of Complex Systems, by Hiroki Sayama (free online textbook)

Things Worth Reading
Everything you wanted to know about Data Analysis and Fitting but were afraid to ask, by Peter Young
Machine Learning, Statistical Inference and Induction Notebook (by Cosma Shalizi)
Power Law distributions, etc. Notebook (by Cosma Shalizi)
Statistics Done Wrong, The woefully complete guide (by Alex Reinhart)
Some Advice on Process for [Research Projects]