DIMACS REU 2012 Schedule

In the MapReduce framework, originally proposed by Google, and its open source implementation, Hadoop, the computation is performed over a set of (key,value) pairs and consists of a map stage, a shuffle stage and a reduce stage. During each round of the computation, these stages are executed sequentially, while data-intensive operations executed by a single stage can be parallelized over many machines. Assuming that the communication cost of the machines and their space usage must be small with respect to the input size, the design of algorithms that minimize the number of rounds becomes quite challenging. This work describes various solutions to the undirected graph connectivity problem and minimum spanning tree, discussing recent results in literature and highlighting open problems. Note: This talk is required only for Czech participants.

The output from numerical simulations and modern experiments consists of large data sets that contain errors (ranging from small to large). This raises the question of how to determine what features are due to errors and what features "really" exist. I will discuss how algebraic topology can be used to address this question.

Abstract TBA.

To promote cultural exchange and socialization, students will present cultural information as a component of the DIMACS/DIMATIA program. Czech students will present information about the Czech Republic, while domestic students will present information about life in the United States or about their own cultural background.

Two panel discussions will introduce students to the graduate school admissions process and answer questions about applying to grad school, as well as graduate student life. One panel will consist of faculty representing REU-related departments; the other will consist of graduate students.

Faculty Panelists: TBA
Student Panelists: TBA