Emily B. Fox is an Assistant Professor in the Department of Statistics at the University of Washington, having joined in 2012 from a prior position at the Wharton Department of Statistics, University of Pennsylvania. From 2009-2011 she was a postdoc in the Duke Statistical Science Department, and received her S.B., M.Eng., E.E. and Ph.D. in EECS at MIT. Her doctoral thesis was awarded the 2009 Leonard J. Savage Thesis Award in Applied Methodology and the 2009 MIT EECS Jin-Au Kong Outstanding Doctoral Thesis Prize. Her research interests include Bayesian nonparametrics, Bayesian dynamic modeling and time series analysis. The work emphasizes methodology for high-dimensional, sparsely sampled data with applications in neuroscience, health monitoring, and finance, amongst others.

Bayesian Dynamic Modeling: Sharing Information Across Time and Space
In this talk we will highlight some of the benefits and challenges associated with harnessing the temporal structure present in many datasets. We focus on Bayesian dynamic modeling approaches, and in particular, the idea of sharing information across time and "space", where space generically refers to the dimensions of the time series.
We exploit nonparametric and hierarchical models to capture repeated patterns in time and similar structure in space, enabling the modeling of complex and high-dimensional time series. Applications of such approaches are quite diverse, and in this talk we will demonstrate this by touching upon our work in the tasks of speaker diarization, analyzing human motion, detecting changes in volatility of stock indices, parsing EEG, word classification from MEG, and predicting rates of violent crimes in DC and influenza rates in the US.

Marina Meila works in computationally efficient statistical learning methods for domains with high dimensions, or with combinatorial or algebraic structure. Currently, she studies the modeling of preference and ranked data, the theoretical foundatons of clustering, and manifold learning. Dr. Meila received her MS from the Polytechnic Institute of Bucharest and her PhD from the Massachusetts Institute of Technology. She is currently Associate Professor of Statistics and Adjunct Associate Professor of Computer Science at the University of Washington, Seattle.

Geometrically preserving manifold learning
In recent years, manifold learning has become increasingly popular as a tool for performing non-linear dimensionality reduction. This has led to the development of numerous algorithms of varying degrees of complexity that aim to recover man ifold geometry using either local or global features of the data.
Building on the Laplacian Eigenmap framework, we propose a new paradigm that off ers a guarantee, under reasonable assumptions, that any manifold learning algori thm will preserve the geometry of a data set. Our approach is based on augmentin g the output of embedding algorithms with geometric information embodied in the Riemannian metric of the manifold. The Riemannian metric allows us to compute ge ometric quantities (such as angle, length, or volume) for any coordinate system or embedding of the manifold. This geometric faithfulness, which is not guarante ed for most algorithms, allows us to define geometric measurements that are inde pendent of the algorithm used, and hence move seamlessly from one algorithm to a nother. In this work, we provide an algorithm for estimating the Riemannian metric from data, consider its consistency, and demonstrate the advantages of our ap proach in a variety of examples.
As an application of this new framework, we develop a new, principled, unsupervised to selecting the scale parameter in manifold learning, based on optimizing the geometric self-consistency w.r.t the scale.
Joint work with Dominique Perrault-Joncas.

Jo-Anne Ting is a member of the Data Mining group at Bosch Research in Palo Alto, CA. She was an NSERC Postdoctoral Fellow from 2010 to June 2011 at the University of British Columbia with Kevin Murphy and did a postdoc in 2009 at the University of Edinburgh with Sethu Vijayakumar. She received her PhD in Computer Science from the University of Southern California in 2009, advised by Stefan Schaal. Prior to that, she graduated with a BASc in Computer Engineering from the University of Waterloo, Canada, in 2003.

Technical Challenges of Big Data Learning: What You Didn't Learn in Graduate School
With the on-going "Big Data" trend in academia and industry, focus is slowly shifting to application domains beyond popular machine learning applications such as product recommendations, advertising and social networking. Energy management, personalized medicine, new materials discovery, drug design and industrial automation are some examples of such application domains with unique and still-to-be addressed research challenges. In this talk, we describe key technical challenges in Big Data learning being addressed at Bosch and showcase opportunities for future research in machine learning and data mining.

Aurelie C Lozano is a research staff member in the Machine Learning group at the IBM T.J. Watson Research Center. She received her Ph.D. from Princeton University, where she was recipient of the Gordon Y.S. Wu Fellowship, the University's highest award for graduate students in engineering and computer science. Aurelie's research interests include machine learning, statistics and data mining. Her current focus is on high dimensional data analysis and predictive modeling, with applications including computational biology, business analytics, and environmental sciences.

Robust Sparse Estimation of Multiresponse Regression and Inverse Covariance Matrix
Robustness is an important aspect often overlooked in the sparse learning literature, while critical when dealing with high dimensional noisy data. The traditional penalized likelihood-based estimators lack resilience with respect to outliers and model misspecification, hence sparse learning methods such as the Lasso break down. In this talk, we will present an alternative approach, that of minimizing regularized distance criteria, which are motivated by the minimum distance functionals used in nonparametric methods for their automatic� robustness. We shall provide theoretical justification for the proposed estimators, and also shed light on their robustness by connecting them to weighted versions of lasso and graphical lasso, where the weights intuitively explain the robustness. We shall illustrate the relevance of our approach through simulations and genomic data analysis. These also confirm that outliers can severely influence the variable selection accuracy of existing sparse learning methods.

Cindi Thompson leads Social and Text Analytics Solution development efforts with Deloitte LLP's Analytics practice. She has over 13 years experience in social media analytics, text analytics, and data mining in areas such as social business strategy, document review systems, sentiment analysis, information extraction, and adaptive recommendation systems. Her R&D and project management experience has been in industrial, consulting, and academic settings. Before coming to Deloitte, Cindi worked four years as a professor then joined PwC for six years as a Research Manager before coming to Deloitte. Cindi has a PhD in Computer Sciences from the University of Texas - Austin.

Social Media Analytics - a Business Perspective
Cindi will discuss her experience as part of a Consulting organization with the emerging field of Social Media Analytics. She will present a business case study and some technical background.