Spring 2017

February 14, 2017 to February 15, 2017

An Information Theoretic Perspective on the ExplorationExploitation Tradeoff

Speaker: Daniel Russo (Northwestern Kellogg School of Management)

An information-theoretic perspective on the exploration/exploitation tradeoff Modern online marketplaces feed themselves: they rely on historical data to optimize content and user-interactions, but it’s the data generated from these...

February 21, 2017 to February 22, 2017

Dynamic Learning in Strategic Pricing Games

Speaker: John Birge (University of Chicago)

In monopoly pricing situations, firms should optimally vary prices to learn demand. The variation must be sufficiently high to ensure complete learning.  In competitive situations, however, varying prices provides information to competitors...

February 28, 2017 to March 1, 2017

Distributed Energy Resource Control and Network Optimization

Speaker: Duncan Callaway (University of California, Berkeley)

In this talk I'll discuss a variety of theory, simulation and experimental work aimed at understanding how distributed energy resources (DERs; such as batteries, flexible loads and photovoltaic generators) impact distribution networks and...

March 28, 2017 to March 29, 2017

H-Infinity Optimal Control on Networks

Speaker: Anders Rantzer (Lund University)

Classical control theory does not scale well for large systems like traffic networks, power networks and chemical reaction networks. However, in this lecture we will present a class of networked control problems for which scalable...

April 4, 2017 to April 5, 2017

Capacity via Symmetry: Extensions and Practical Consequences

Speaker: Henry Pfister (Duke University)

Recently, sequences of error-correcting codes with doubly-transitive permutation groups were shown to achieve capacity on erasure channels under symbol-wise maximum a posteriori (MAP) decoding. From this, it follows that Reed-Muller and...

April 6, 2017 to April 7, 2017

A System Level Approach to Controller Synthesis

Speaker: Nikolai Matni (California Institute of Technology)

Biological and advanced cyberphysical control systems often have limited, sparse, uncertain, and distributed communication and computing in addition to sensing and actuation. Fortunately, the corresponding plants and performance requirements...

April 11, 2017 to April 12, 2017

Geometries of Word Embeddings

Speaker: Pramod Viswanath (University of Illinois Urbana-Champaign)

Real-valued word vectors have transformed NLP applications; popular examples are word2vec and GloVe, recognized for their ability to capture linguistic regularities via simple geometrical operations. In this talk, we demonstrate further...

April 13, 2017 to April 14, 2017

Data-Driven Models in Power Systems

Speaker: Ram Rajagopal (Stanford University)

Increase in supply-side variability due to increases in renewable generation requires integrating new resources utilizing improved models of the power system to reduce electricity delivery costs. Data from consumers and markets has become...

April 19, 2017

The Landscape of Some Statistical Problems

Speaker: Andrea Montanari (Stanford University)

Most high-dimensional estimation and prediction methods propose to minimize a cost function (empirical risk) that is written as a sum of losses associated to each data point (each example). Studying the landscape of the empirical risk is...

April 27, 2017 to April 28, 2017

Non-Convex Learning via Stochastic Gradient Langevin Dynamics

Speaker: Maxim Raginsky (University of Illinois)

Stochastic Gradient Langevin Dynamics (SGLD) is a popular variant of Stochastic Gradient Descent, where properly scaled isotropic Gaussian noise is added to an unbiased estimate of the gradient at each iteration. This modest change allows...

May 16, 2017 to May 17, 2017

Stable Optimal Control and Semicontractive Dynamic Programming

Speaker: Dimitri Bertsekas (Massachusetts Institute of Technology)

We consider discrete-time infinite horizon deterministic optimal control problems with nonnegative cost, and a destination that is cost-free and absorbing. The classical linear-quadratic regulator problem is a special case. The analysis aims...