2017, Exams of Algorithms and Programming

Download 2017 and more Exams Algorithms and Programming in PDF only on Docsity!

JANUARY 30

9:20am Introduction Mathieu Desbrun Caltech

9:30am Data Privacy: How to Survive the Inference Avalanche

Reza Shokri Cornell University

10:30am Coffee Break

11:00am Data-Driven Probabilistic Modeling and High-Performance Computing: Algorithms and Applications to Physical and Biological Systems

Paris Perdikaris Massachusetts Institute of Technology

1:15pm Extreme Events and Metastability in Fluids and Waves

Tobias Grafke New York University

2:15pm Coffee Break

2:45pm Towards a Theory of Fairness in Machine Learning

Jamie Morgenstern University of Pennsylvania

3:45pm Coffee Break

4:15pm Shared-Memory Parallelism Can Be Simple, Fast, and Scalable

Julian Shun UC Berkeley

ALL TALKS ARE ONE HOUR LONG, AND WILL BE HELD IN ANNENBERG 105.

COFFEE BREAKS WILL BE HELD IN THE ANNENBERG LOBBY.

Towards a Theory of Fairness in Machine Learning

Jamie Morgenstern University of Pennsylvania

Algorithm design has moved from being a tool used exclusively for designing systems to one used to present people with personalized content, advertisements, and other economic opportunities. Massive amounts of information is recorded about people's online behavior including the websites they visit, the advertisements they click on, their search history, and their IP address. Algorithms then use this information for many purposes: to choose which prices to quote individuals for airline tickets, which advertisements to show them, and even which news stories to promote. These systems create new challenges for algorithm design. When a person's behavior influences the prices they may face in the future, they may have a strong incentive to modify their behavior to improve their long-term utility; therefore, these algorithms' performance should be resilient to strategic manipulation. Furthermore, when an algorithm makes choices that affect people's everyday lives, the effects of these choices raise ethical concerns such as whether the algorithm's behavior violates individuals' privacy or whether the algorithm treats people fairly.

Machine learning algorithms in particular have received much attention for exhibiting bias, or unfairness, in a large number of contexts. In this talk, I will describe my recent work on developing a definition of fairness for machine learning. One definition of fairness, encoding the notion of 'fair equality of opportunity', informally, states that if one person has higher expected quality than another person, the higher quality person should be given at least as much opportunity as the lower quality person. I will present a result characterizing the performance degradation of algorithms which satisfy this condition in the contextual bandits setting. To complement these theoretical results, I then present the results of several empirical evaluations of fair algorithms.

I will also briefly describe my work on designing algorithms whose performance guarantees are resilient to strategic manipulation of their inputs, and machine learning for optimal auction design.

Biography:

Jamie Morgenstern is a Warren Center postdoctoral fellow in Computer

Science and Economics at the University of Pennsylvania. She received her

Ph.D. in Computer Science from Carnegie Mellon University in 2015, and

her B.S. in Computer Science and B.A. in Mathematics from the University of

Chicago in 2010. Her research focuses on machine learning for mechanism

design, fairness in machine learning, and algorithmic game theory. She

received a Microsoft Women's Research Scholarship, an NSF Graduate

Research Fellowship, and a Simons Award for Graduate Students in

Theoretical Computer Science.

Data-Driven Probabilistic Modeling and High-Performance Computing: Algorithms and Applications to Physical and Biological Systems

Paris Perdikaris

Massachusetts Institute of Technology

The analysis of complex physical and biological systems necessitates the accurate

resolution of interactions across multiple spatio-temporal scales, the consistent

propagation of information between concurrently coupled multi-physics processes, and

the effective quantification of model error and parametric uncertainty. Addressing these

grand challenges is a multi-faceted problem that poses the need for a highly sophisticated

arsenal of tools in stochastic modeling, high-performance scientific computing, and

probabilistic machine learning. Through the lens of three realistic large-scale applications,

this talk aims to demonstrate how the compositional synthesis of such tools is introducing

a new paradigm in scientific discovery. First, we present multi-scale blood flow simulations

in the human brain, and show how high-order methods, massively parallel computing, and

concurrent coupling of multi-physics solvers can uncover intrinsic physiological

mechanisms in health and disease. We will demonstrate how the introduction of

probabilistic machine learning techniques, and the key concept of multi-fidelity modeling,

provide a scalable platform for information fusion and lead to significant computational

expediency gains. The second application involves an environmental study that illustrates

how machine learning tools enable the synergistic combination of simulations, noisy

measurements and empirical models towards quantifying the anthropogenic effect in the

increasing acidification of coastal waters, and developing a cost-effective monitoring and

prediction mechanism. Lastly, we consider the shape optimization of super-cavitating

hydrofoils of an ultrafast marine vessel for special naval operations. Specifically, we show

how the combination of turbulent multi-phase flow simulations and the concept of multi-

fidelity Bayesian optimization allows us to tackle complex engineering design problems in

which a rigorous assessment of uncertainty and risk becomes critical in policy and decision

making.

Biography: Paris Perdikaris received his PhD in Applied Mathematics from Brown University in May 2015. His expertise lies in probabilistic machine learning, computational fluid dynamics, multi-fidelity modeling, uncertainty quantification, and parallel scientific computing. While at Brown he developed scalable machine learning algorithms for predictive multi-fidelity modeling of high- dimensional systems. A parallel research thrust involved developing mathematical models for simulating cardiovascular fluid flows and assessing the characteristics of cerebral pathologies such as cerebral aneurysms in-silico. In June 2015, he moved to MIT as a post-doctoral research associate at the department of Mechanical Engineering and the MIT Sea Grant College Program. His research at MIT is focused on designing a scalable data-driven framework for uncertainty quantification, inverse problems, design optimization, and beyond. The developed algorithms are currently used for risk-averse design optimization of super-cavitating hydrofoils, as well as data assimilation of noisy measurements in coastal regions. Moreover, he has co-advised an undergraduate student working on active learning and data acquisition under uncertainty, and a masters student working on deep learning techniques for object recognition, tracking, and autonomous marine navigation. From 2010-present he has been actively involved in several research projects funded by major US agencies including DOE, AFOSR, NIH and DARPA.

Shared-Memory Parallelism

Can Be Simple, Fast, and Scalable

Julian Shun UC Berkeley

Parallelism is the key to achieving high performance in computing. However,

writing efficient and scalable parallel programs is notoriously difficult, and often

requires significant expertise. To address this challenge, it is crucial to provide

programmers with high-level tools to enable them to develop solutions more

easily, and at the same time emphasize the theoretical and practical aspects of

algorithm design to allow the solutions developed to run efficiently under many

possible settings. My research addresses this challenge using a three-pronged

approach consisting of the design of shared-memory programming techniques,

frameworks, and algorithms for important problems in computing. In this talk, I

will present tools for deterministic parallel programming, large-scale shared-

memory algorithms that are efficient both in theory and in practice, and Ligra, a

framework for simplifying the programming of shared-memory graph algorithms.

Biography:

Julian Shun is currently a Miller Research Fellow (post-doc) at UC Berkeley. He

obtained his Ph.D. in Computer Science from Carnegie Mellon University, and

his undergraduate degree in Computer Science from UC Berkeley. He is

interested in developing large-scale parallel algorithms for graph processing,

and parallel text algorithms and data structures. He is also interested in

designing methods for writing deterministic parallel programs and

benchmarking parallel programs. He has received the ACM Doctoral

Dissertation Award, CMU School of Computer Science Doctoral Dissertation

Award, Miller Research Fellowship, Facebook Graduate Fellowship, and a best

student paper award at the Data Compression Conference.