<< February 2018 >>

Thursday, February 1, 2018

Seminar 217, Risk Management: Interpretable proximate factors for large dimensions

Seminar: Risk Seminar | February 1 | 12:30-2 p.m. | 1011 Evans Hall

 Speaker: Markus Pelger, Stanford

 Center for Risk Management Research

This papers deals with the approximation of latent statistical factors with sparse and easy-to-interpret proximate factors. Latent factors in a large-dimensional factor model can be estimated by principal component analysis, but are usually hard to interpret. By shrinking the factor weights, we obtain proximate factors that are easier to interpret. We show that proximate factors consisting of...   More >

Wednesday, February 7, 2018

Large deviations for two time-scale jump-diffusions and Markov chain models

Seminar: Probability Seminar | February 7 | 3:10-4 p.m. | 1011 Evans Hall

 Lea Popovic, Concordia University

 Department of Statistics

For a number of processes in biology the appropriate stochastic modelling is done in
terms of multi-scale Markov processes with fully dependent slow and fast fluctuating variables.
The most common examples of such multi-scale processes are deterministic evolutions, jump-diffusions,
and state dependent Markov chains. The law of large numbers limit, central limit theorem,
and the corresponding...   More >

Applied Statistics at Tesla

Seminar: Neyman Seminar | February 7 | 4-5 p.m. | 1011 Evans Hall

 Swupnil Sahai, Tesla; Andrej Karpathy, Tesla

 Department of Statistics

From estimating the time to failure of battery modules for Reliability Engineering to predicting lane lines from images for Autopilot, statistics plays a vital role in building all of Tesla’s products. In this talk, we present the ways in which Tesla is changing the future of sustainable energy and discuss how statisticians will help us get there.

Wednesday, February 14, 2018

Scaling limits for percolated random planar maps

Seminar: Probability Seminar | February 14 | 3:10-4 p.m. | 1011 Evans Hall

 Nina Holden, Concordia University

 Department of Statistics

The Schramm-Loewner evolution (SLE) is a family of random fractal curves, which is the proven or conjectured scaling limit of a variety of two-dimensional lattice models in statistical mechanics. Liouville quantum gravity (LQG) is a model for a random surface which is the proven or conjectured scaling limit of discrete surfaces known as random planar maps (RPM). We prove scaling limit results for...   More >

Thursday, February 15, 2018

Seminar 217, Risk Management: Digitally-driven change in the insurance industry—disruption or transformation?

Seminar: Risk Seminar | February 15 | 12:30-2 p.m. | 1011 Evans Hall

 Speaker: Jeffrey Bohn, Swiss Re

 Center for Risk Management Research

As technology continues to insinuate itself into all facets of financial services, the insurance industry faces a slow-motion parade of promise, possibilities, prematurity, and pared-down expectations. Digitization, the birth of InsurTech, machine intelligence, and the collection & curation of (orders of magnitude) more structured & unstructured data are changing (and will continue to change) the...   More >

Testing for two-stage experiments in the presence of interference

Seminar: Neyman Seminar | February 15 | 4-5 p.m. | 1011 Evans Hall

 Guillaume Basse, Harvard University

 Department of Statistics

Many important causal questions concern interactions between units, also known as interference. Examples include interactions between individuals in households, students in schools, and firms in markets. Standard analyses that ignore interference can often break down in this setting: estimators can be badly biased, while classical randomization tests can be invalid. In this talk, I present recent...   More >

Wednesday, February 21, 2018

Low-temperature localization of directed polymers

Seminar: Probability Seminar | February 21 | 3:10-4 p.m. | 1011 Evans Hall

 Erik Bates, Stanford University

 Department of Statistics

On the d-dimensional integer lattice, directed polymers are paths of a random walk that have been reweighted according to a random environment that refreshes at each time step. The qualitative behavior of the system is governed by a temperature parameter; if this parameter is small, the environment has little effect, meaning all possible paths are close to equally likely. If the parameter is made...   More >

Low-temperature localization of directed polymers

Seminar: Probability Seminar | February 21 | 3:10-4 p.m. | 1011 Evans Hall

 Erik Bates, Stanford University

 Department of Statistics

On the d-dimensional integer lattice, directed polymers are paths of a random walk that have been reweighted according to a random environment that refreshes at each time step. The qualitative behavior of the system is governed by a temperature parameter; if this parameter is small, the environment has little effect, meaning all possible paths are close to equally likely. If the parameter is made...   More >

Weina Wang- Delay Bounds And Asymptotics In Cloud Computing Systems

Seminar: Other Related Seminars | February 21 | 3:30-5 p.m. | 3110 Etcheverry Hall

 Weina Wang, Illinois Urbana-Campaign

 Industrial Engineering & Operations Research

With the emergence of big-data technologies, cloud computing systems are growing rapidly in size and becoming more and more complex, making it costly to conduct experiments and simulations. Therefore, modeling computing systems and characterizing their performance analytically are more critical than ever in identifying bottlenecks, informing system design, and facilitating provisioning.

Recent Advances in Algorithmic High-Dimensional Robust Statistics

Seminar: Neyman Seminar | February 21 | 4-5 p.m. | 1011 Evans Hall

 Ilias Diakonikolas, USC

 Department of Statistics

Fitting a model to a collection of observations is one of the quintessential problems in machine learning. Since any model is only approximately valid, an estimator that is useful in practice must also be robust in the presence of model misspecification. It turns out that there is a striking tension between robustness and computational efficiency. Even for the most basic high-dimensional tasks,...   More >

Thursday, February 22, 2018

Seminar 217, Risk Management: Solving the “curse of dimensionality” problem in multi-asset-class risk models

Seminar: Risk Seminar | February 22 | 12:30-2 p.m. | 1011 Evans Hall

 Speaker: Jose Menchero, Bloomberg

 Center for Risk Management Research

Estimating a robust risk model risk for a portfolio that spans multiple asset classes is a challenging task due to the “curse of dimensionality” (i.e., the problem of estimating too many relationships from too few observations). While the sample covariance matrix is easily computed, it is susceptible to capturing spurious relationships that make it unsuitable for portfolio construction purposes....   More >

Dr. Aaron McKenna, Department of Genome Sciences, University of Washington: Resolving whole organism cell fate with CRISPR/Cas9

Seminar: Other Related Seminars | February 22 | 4-5 p.m. | Soda Hall, HP Auditorium 306

 Center for Computational Biology, Electrical Engineering and Computer Sciences (EECS)

Abstract: Multicellular organisms develop by way of a lineage tree, a series of cell divisions that give rise to cell types, tissues, and organs. However, our knowledge of the cell lineage and its determinants remains extremely fragmentary for nearly all species. This includes all vertebrates and arthropods such as Drosophila, wherein cell lineage varies between individuals; embryos and organs.

Monday, February 26, 2018

Dr. Mingfu Shao, Department of Computational Biology, Carnegie Mellon University: Efficient algorithms for large-scale transcriptomics and genomics

Seminar: Other Related Seminars | February 26 | 4-5 p.m. | Soda Hall, HP Auditorium 306

 Center for Computational Biology, Electrical Engineering and Computer Sciences (EECS)

Title:


Abstract:

I will present modeling and algorithmic designs for two challenging problems in biology and argue that efficient computational methods enable significant advances in our understanding of cell machinery and genome evolution. The first problem is the assembly of full-length transcripts -- the collection of expressed gene products in cells -- from noisy and highly...   More >

GraphXD Seminar: Vector Representations of Graphs and the Maximum Cut Problem

Seminar: Other Related Seminars | February 26 | 4-5:30 p.m. | 1011 Evans Hall

 David P. Williamson, Operations Research and Information Engineering, Cornell University

 Berkeley Institute for Data Science

In this talk, I will look at a classical problem from graph theory of finding a large cut in a graph. We’ll start with a 1967 result of Erdős that showed that picking a random partition of the graph finds a cut that is at least half the largest possible cut. We’ll then describe a result due to Goemans and myself from 1995 that shows that by representing the graph as a set of vectors, one per...   More >

Wednesday, February 28, 2018

Markovian Solutions to Scalar Conservation Law

Seminar: Probability Seminar | February 28 | 3:10-4 p.m. | 1011 Evans Hall

 Fraydoun Rezakhanlou, U C Berkeley

 Department of Statistics

According to a classical result of Bertoin (1998), if the initial data for Burgers equation is a Levy Process with no positive jump, then the same is true at later times and there is an explicit equation for the evolution of the associated Levy measures. In 2010, Menon and Srinivasan published a conjecture for the statistical structure of solutions to scalar conservation laws with certain Markov...   More >

Algorithmic Regularization in Over-parameterized Matrix Recovery and Neural Networks with Quadratic Activations

Seminar: Neyman Seminar | February 28 | 4-5 p.m. | 1011 Evans Hall

 Tengyu Ma, Facebook AI Research

 Department of Statistics

Over-parameterized models are widely and successfully used in deep learning, but their workings are far from understood. In many practical scenarios, the learned model generalizes to the test data, even though the hypothesis class contains a model that completely overfits the training data and no regularization is applied.

In this talk, we will show that such phenomenon occurs in...   More >