Watch Now: Faculty Vision Talks

Eric Darve, Associate Professor of Mechanical Engineering

Solving linear systems is at the core of many if not most scientific and engineering calculations.  With the advent of large parallel supercomputers and big-data problems with terabytes of data, we are now dealing with matrices that are both very large in size and have a complex structure (i.e., distribution of their eigenvalues). As a result, many conventional linear solvers fail. For example, the convergence of iterative methods often degrades with problem size and is found toplateaus for large enough problems. We will present a new generation of linear solvers designed to solve especially difficult linear problems, because of the size or the conditioning of the matrix.

 

Ramesh Johari, Associate Professor of Management Science and Engineering

I'll describe SOAL, a lab that I co-direct along with four other faculty.  Our lab's focus is at theinterface of two macroscopic trends: (1) tremendous advances in information andcommunication technology; and (2) ever-richer networked social and economic interactions.  The first trend has enabled the second (e.g., online social networks); and the data generated from thesecond amplifies the first (e.g., recommendation systems and collaborative filtering).  We focuson research problems that arise where these two trends meet.  I'll overview some currentprojects in the lab, with particular emphasis on market design, online recommendations, anddigital experimentation.

 

Ron Dror, Associate Professor of Computer Science and ICME

Professor Dror will discuss computational approaches to create comprehensive models of biological structure, spatial organization, and dynamics, from the atomic scale to the cellular scale.  This requires combining several types of simulation with diverse sources of experimental data.

 

Machine Learning in general and deep learning in particular have had an enormous impact in many fields, including computer vision and speech recognition.  These approaches are making animpact in computational chemistry and computational biology as well.  I will discuss my vision for bringing these fields together and where the biggest impacts will be, highlighting advances in fundamental biophysics and drug design.

 

 

Jack Poulson, Assistant Professor of Mathematics and ICME

Despite the current surge in academic and commercial interest in first-order methods foroptimization, research on sparse-direct techniques for distributed second-order methods and (generalized) least squares solvers has received comparatively little attention. In this talk, Professor Poulson will briefly describe his ongoing efforts in building a high-quality, open-sourcelibrary for distributed-memory linear algebra and (as of recently) optimization.

 

Alison Marsden, Associate Professor of Pediatrics and ICME

I will highlight a range of computational tools for cardiovascular bioengineering, including patient specific modeling, uncertainty quantification, multiscale modeling, and optimization.   I will then give an overview of how these tools are being applied in real clinical applications to impact thetreatment of cardiovascular disease, particularly for children born with severe congenital heart defects.   I will then discuss my vision for future directions in cardiovascular blood flow simulations and clinical application.  

 

Chris Re, Assistant Professor of Computer Science

Many pressing questions in science are macroscopic, as they require scientists to integrate information from numerous data sources, often expressed in natural languages or in graphics; these forms of media are fraught with imprecision and ambiguity and so are difficult for machines to understand. I describe DeepDive, which is a new type of system designed to cope with these problems. It combines extraction, integration and prediction into one system. For some paleobiology and materials science tasks, DeepDive-based systems have surpassed human volunteers in data quantity and quality (recall and precision). DeepDive is also used by scientists in areas including genomics and drug repurposing, by a number of companies involved in various forms of research, and by law enforcement in the fight against human trafficking. DeepDive does not allow users to write algorithms; instead, it asks them to write only features. A key technical challenge is scaling up the resulting inference and learning engine, and I will describe our line of work in computing without using traditional synchronization methods including Hogwild! andDimmWitted. DeepDive is open source on github and available from DeepDive.stanford.edu.

 

Lexing Ying, Professor of Mathematics and ICME

This autumn, ICME will welcome our first cohort of Masters students who will study under a brand-new specialized degree track: Imaging Sciences.  Professor Ying will talk about this new program.