[Cs-affiliates] CS 591 topics courses descriptions - Fall 2016

Devits, Christopher R cdevits at bu.edu
Tue Aug 23 11:51:09 EDT 2016


Dear CS Students,

We wanted to pass along updated information about the different CS topics courses (those listed as CS 591) that will be offered this upcoming Fall. Note there is one new addition since our last email – CS 591 E2 Optimization Methods & their Applications to be taught by one of new faculty members, Dr. Alina Ene. The expanded description is below, and it is now up on the university class schedule. Please register for it if you’re interested.


·         CAS CS 591 E2 Optimization Methods & their Applications


Course Description: At their core, many real-world problems involve optimization problems that are complex and challenging, and they require principled mathematical and algorithmic solutions. Companies such as Google and Facebook use powerful algorithmic primitives such as gradient descent to solve a host of challenging problems. These primitives in turn are based on a beautiful mathematical theory developed in the context of convex and discrete optimization. In this class, we will explore several of these optimization primitives that have found significant applications and we will develop a mathematical toolkit that can be used to tackle real-world problems. Subject matter: The main topics covered include convex optimization methods and their applications, such as gradient descent, mirror descent, the multiplicative weights update method, the Frank-Wolfe gradient descent algorithm; discrete and combinatorial optimization methods and their applications, such as submodular functions, submodular minimization via the ellipsoid method and gradient descent, and submodular maximization via the Greedy algorithm; and algorithms for linear algebra and their applications, such as the power method, singular value decompositions, principal component analysis, spectral clustering, and regression. Prerequisites: The class will assume familiarity with fundamental algorithmic problems and techniques, and a working knowledge of calculus and linear algebra. A strong undergraduate algorithms class, CS 330 or equivalent is required. Students should also have had either a second level algorithms class (CS 530 at BU) or some other class(es) which center on more advanced algorithmic methods. The course evaluation will be based on homework assignments, a final project, and lecture notes scribing.

Below you will find expanded titles and descriptions for all our 591 courses. Please feel free to contact the instructor or the department with any questions. Several are closed and we are still managing wait lists for those through September 19th (the last day to add a course).

Enjoy the final few days of the summer break, and we’ll see you all soon!

Best,

-chris

------------
Christopher DeVits
Department Administrator
Computer Science

——————
CAS CS 591 B1
Networks and Markets: Theory and Applications

Course Description: Recently, the concept of a network has outgrown the narrower engineering mindset of a collection of interconnected machines to become more broadly relevant in a variety of applied settings which feature connectedness.  Biological networks, social networks, online advertising networks, and networks involving hyperlinks, i.e., the WWW, are all examples of domains in which the theory and practice of networking science has now being applied.  In parallel with this trend is the rise of online markets as a mediation point for commercial activity.  Beginning with the advent of Internet platforms like eBay that employ online auctions, are many fascinating new markets: pay-per-click advertising markets, prediction markets, and two-sided platforms such as Uber and Airbnb.  All of these application domains draw deeply on established theory that is often unfamiliar territory to computer scientists:  auction design, mechanism design, and the theory of matching markets.  Finally, many important new technologies incorporate a mix of networks and markets, including information networks and recommender systems.  In this class, we will build on the highly successful undergraduate text of Easley and Kleinberg to learn about the underlying theory of networks and markets, and understand how modern-day applications connect to these foundations. Prerequisites: This course is designed for declared CS majors who are fulfilling 400-level electives, as well as Masters students and Ph.D. students.  CS 112, CS 131, and some knowledge of probability and statistics is required.

Instructor: John Byers

——————
CAS CS 591 C1
Data Science with Python


Course Description: The course emphasizes practical skills in working with data, while introducing students to a wide range of techniques that are commonly used in the analysis of data, such as clustering, classification, regression, and network analysis. The goal of the class is to provide to students a hands-on understanding of classical data analysis techniques and to develop proficiency in applying these techniques in a modern programming language (Python).  Lectures will present the fundamentals of each technique; focus is not on the theoretical underpinnings of the methods, but rather on helping students understand the practical settings in which these methods are useful. Class discussion will study use cases and will go over relevant Python packages that will enable the students to perform hands-on experiments with their data.  Note this class is different from CS 565 (Data Mining): while CS 565 focuses on the fundamental algorithmic problems around a set of data-mining problems and emphasizes on the analysis of the algorithms for certain data analysis tasks, this class will focus on how these algorithms woork in practice.  Target audience: This course is targeted towards graduate or advanced undergraduate students who need to be proficient on working and analyzing large datasets for their research or aim to find a job that will require data-analysis skills.



Instructor: Mark Crovella



——————
CAS CS 591 C2
Theory of Compressive Sensing in the Age of Big Data

Course Description: One of the most exciting developments in this age of big data, along with our deeper understanding of signal processing, machine learning, data fusion, etc., is our deeper understanding of sparsity. Indeed, sparsity has become a very important concept in recent years with the key idea being that many classes of natural signals can be described by only a small number of significant degrees of freedom. This course offers a complete coverage of the recently emerged field of compressed sensing, which asserts that, if the true signal is sparse to begin with, accurate, robust, and even perfect signal recovery can be achieved from just a few randomized measurements. This idea has resulted in novel devices such as single-pixel camera, better MRI scanner, more accurate face object recognition algorithms, more efficient communication protocols, etc. The focus is on describing the novel ideas that have emerged in sparse recovery with emphasis on theoretical foundations, practical numerical algorithms, and various related applications in image recovery, machine learning, neural signal reconstruction, etc. Students from diverse background (ECE, BME, MechE, Medical School, etc) who are either interested in the subject or want to apply this new theory for their research are encouraged to attend.

Instructor: Peter Chin

——————

CAS CS 591 E1
Mobile Application Development

Course Description: Students will utilize agile software engineering practices in this hands-on course to design and implement mobile applications using Java and the Android SDK.  Students will initially implement several small mobile applications utilizing core android technologies, after which students will be grouped into small groups, collaborating on a larger final project. Topics will include UI development, action bars, multi-touch, gestures, database and file I/O.  Students will also learn to make rich applications by consuming location and sensor information from device hardware. No previous mobile application development experience is required, but a strong understanding of object-oriented programming and database development (from CS 112 and CS 460 or equivalent) is necessary.

Instructor: Shereif El-Sheikh

——————

CAS CS 591 E2
Optimization Methods & their Applications

Course Description: At their core, many real-world problems involve optimization problems that are complex and challenging, and they require principled mathematical and algorithmic solutions. Companies such as Google and Facebook use powerful algorithmic primitives such as gradient descent to solve a host of challenging problems. These primitives in turn are based on a beautiful mathematical theory developed in the context of convex and discrete optimization. In this class, we will explore several of these optimization primitives that have found significant applications and we will develop a mathematical toolkit that can be used to tackle real-world problems. Subject matter: The main topics covered include convex optimization methods and their applications, such as gradient descent, mirror descent, the multiplicative weights update method, the Frank-Wolfe gradient descent algorithm; discrete and combinatorial optimization methods and their applications, such as submodular functions, submodular minimization via the ellipsoid method and gradient descent, and submodular maximization via the Greedy algorithm; and algorithms for linear algebra and their applications, such as the power method, singular value decompositions, principal component analysis, spectral clustering, and regression. Prerequisites: The class will assume familiarity with fundamental algorithmic problems and techniques, and a working knowledge of calculus and linear algebra. A strong undergraduate algorithms class, CS 330 or equivalent is required. Students should also have had either a second level algorithms class (CS 530 at BU) or some other class(es) which center on more advanced algorithmic methods. The course evaluation will be based on homework assignments, a final project, and lecture notes scribing.

Instructor: Alina Ene

——————

CAS CS 591 L1

Data Mechanics for Pervasive Systems and Urban Applications



Course Description: This course is about how data can move through institutions and computational infrastructures to inform decisions and operations (often in real time) within large systems such as cities that contain a variety of widely distributed, embedded sensors and devices (e.g., sensor networks and smart power grids). The course covers and teaches students to apply some of the tools and methods that computer science and computational thinking provide for facilitating data collection, retrieval, integration, and interpretation in application areas like urban informatics and distributed systems. Students will learn how to use the relational and MapReduce paradigms to assemble analysis, optimization, and decision-making algorithms that can scale up to handle large amounts of static or streaming data. Formal techniques for modeling and ensuring predictable, reliable, and correct behavior of such algorithms, as well as techniques for tracking data provenance, will also be presented. Other potential topics include crowdsourcing, socio-adaptive systems, and modern tools for building offline and online visualizations. Students will apply the tools and methods presented to build platforms and applications that work with real data sets dealing with aspects of urban environments such as mobility (e.g., walkability), employment, traffic and parking, emissions, energy consumption, public safety, and others.



Instructor: Andrei Lapets



——————

CAS CS 591 O1

Iterative Methods in Graph Algorithms and Network Analysis



Course Description: Iterative methods allow us to trade-off precision and running time to produce theoretically sound, fast algorithms for many problems in Convex Optimization. These methods rely on a very “continuous” interpretation of the problem, incorporating techniques such as steepest-descent and coordinate descent that strongly rely on geometric intuition.  In this class, we will turn this perspective on his head and show that iterative methods can be extremely useful in understanding and solving fundamental combinatorial problems on discrete structures such as graphs and networks (i.e. directed graphs). In particular, iterative methods can help us leverage the combinatorial structure of graphs to its fullest and obtain simple and extremely fast algorithms for important primitives like shortest paths, maximum flow and graph partitioning. Many of these novel algorithms run in time essentially linear in  the size of the graph. This has important implications at a time when input graphs may have even billions of nodes and any super-linear running time is practically infeasible.  The course will slowly build up towards these recent, exciting applications. In the process, we will review and develop a number of tools in Convex and Combinatorial Optimization that play an important role in the modern design of graph and network algorithms. While the class will not depend heavily on background material, it will require familiarity with fundamental algorithmic problems and techniques at the level of CS330 or equivalent. A working knowledge of basic linear algebra is also required. Do not hesitate to contact me if you are not sure whether your previous preparation will allow you to benefit from the class.  I expect to cover a large subset of the following topics, probably in a different order from the one given here. In particular, “continuous” and “discrete” topics will be interleaved in a way that a highlights the synergy between these two approaches: — CONTINUOS TOPICS: Review of Linear Programming (LP), Semidefinite Programming (SDP) and Lagrangian Duality.  First-order methods in convex optimization: gradient descent, dual averaging, primal-dual methods First-order methods for large data: coordinate descent and accelerated coordinate descent Interior-point methods



Instructor: Lorenzo Orrechia
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://cs-mailman.bu.edu/pipermail/cs-affiliates/attachments/20160823/339a84ff/attachment-0001.html>


More information about the Cs-affiliates mailing list