What are the opening hours?

Monday 8am - 5pm Tuesday 8am - 5pm Wednesday 8am - 5pm Thursday 8am - 5pm Friday 8am - 5pm

https://www.facebook.com/206384886053286/

Yale Department of Statistics and Data Science

Home
United States
New Haven, CT
Yale Department of Statistics and Data Science

The official page for the Department of Statistics and Data Science at Yale University. and Ph.D.

The Department of Statistics and Data Science at Yale University offers an undergraduate major, M.A. programs, and a newly-created undergraduate certificate in data science. We have active research programs in statistical information theory, statistical genetics and bioinformatics, Bayesian methods, statistical computing, graphical methods, model selection, asymptotics, and other topics. Our facul

ty and students are also active in collaborative research with other departments throughout the university, including computer science, biological sciences, social sciences, physical sciences, engineering, bioinformatics, economics, and applied mathematics. We occupy the James Dwight Dana House, which is listed as a National Historic Landmark and carries the Landmark Plaque of the New Haven Preservation Trust. In addition to office and classroom spaces, the Dana House is home to a computing laboratory for use by our faculty and students.

03/01/2020

Adji Bousso Dieng from Columbia presenting tomorrow at 17 Hillhouse!

Information and Abstract:
Deep learning (DL) is a powerful approach to modeling complex and large scale data. However, DL models lack interpretable quantities and calibrated uncertainty. In contrast, probabilistic graphical modeling (PGM) provides a framework for formulating an interpretable generative process of data and a way to express uncertainty about what we do not know. How can we develop machine learning methods that bring together the expressivity of DL with the interpretability and calibration of PGM to build flexible models endowed with an interpretable latent structure that can be fit efficiently? I call this line of research deep probabilistic graphical modeling (DPGM). In this talk, I will discuss my work on developing DPGM for text data. In particular, I will show how DPGM enables flexible and interpretable topic modeling at large scale, unlocking several known challenges. Furthermore, I will describe how we can account for both local and long-range context, under the DPGM framework, to build a flexible sequential document model that leads to state-of-the-art performance on a downstream document classification task.

Bio: Adji Bousso Dieng is a PhD Candidate at Columbia University where she is jointly advised by David Blei and John Paisley. Her research is in Artificial Intelligence and Statistics, bridging probabilistic graphical models and deep learning.
Dieng is supported by a Dean Fellowship from Columbia University. She won a Microsoft Azure Research Award and a Google PhD Fellowship in Machine Learning. She was recognized as a rising star in machine learning by the University of Maryland.
Prior to Columbia, Dieng worked as a Junior Professional Associate at the World Bank. She did her undergraduate studies in France where she attended Lycee Henri IV and Telecom ParisTech–France’s Grandes Ecoles system. She spent the third year of Telecom ParisTech’s curriculum at Cornell University where she earned a Master in Statistics.

02/25/2020

Berk Ustun from Harvard presenting tomorrow at 17 Hillhouse!

Information and Abstract:
Machine learning is now a general-purpose technology. In many domains, we can build models to support important decisions or automate routine tasks. Yet we may not reap their benefits due to disuse, or even inflict harm due to misuse. In this talk, I will present methodological advances that address these “last mile” challenges. First, I will describe a method to learn simple risk scores that are readily adopted for medical decision support, and discuss applications to adult ADHD diagnosis and ICU seizure prediction. Next, I will describe how machine learning models may harm individuals in consumer-facing applications by violating their right to autonomy. I will then introduce the notion of “recourse” and formalize methods to prevent such harms without interfering in model development.

Bio: Berk Ustun is a postdoc at the Harvard Center for Research on Computation and Society. His research interests are in machine learning, optimization, and human-centered design. In particular, he focuses on developing methods to promote the adoption and responsible use of machine learning in domains such as medicine, consumer finance, and criminal justice.

Berk has built machine learning systems that are now used by major healthcare providers for hospital readmissions prediction, ICU seizure prediction, and adult ADHD screening. His work has been covered by various media outlets, including NPR and Wired, and has won major awards, including the INFORMS Informative Applications in Analytics Award in 2016 and 2019, and the INFORMS Computing Society Best Student Paper.

Berk holds a PhD in Electrical Engineering and Computer Science from MIT, an MS in Computation for Design and Optimization from MIT, and BS degrees in Operations Research and Economics from UC Berkeley.

For links to papers, videos, and software, see: https://www.berkustun.com.

02/23/2020

Emma Pierson from Stanford presenting tomorrow at 17 Hillhouse!

Information and Abstract:
I will describe how to use data science methods to understand and reduce inequality in two domains: criminal justice and healthcare. First, I will discuss how to use Bayesian modeling to detect racial discrimination in policing. Second, I will describe how to use machine learning to explain racial and socioeconomic inequality in pain.

Bio: Emma Pierson is a PhD student in Computer Science at Stanford, supported by Hertz and NDSEG Fellowships. Previously, she completed a master’s degree in statistics at Oxford on a Rhodes Scholarship. She develops statistical and machine learning methods to study two deeply entwined problems: reducing inequality and improving healthcare. She also writes about these topics for broader audiences in publications including The New York Times, The Washington Post, FiveThirtyEight, and Wired. Her work has been recognized by best paper (AISTATS 2018), best poster (ICML Workshop on Computational Biology), and best talk (ISMB High Throughput Sequencing Workshop) awards, and she has been named a Rising Star in EECS and Forbes 30 Under 30 in Science.

02/16/2020

Yixin Wang from Columbia presenting tomorrow at 17 Hillhouse!

Information and Abstract:
Causal inference from observational data is a vital problem, but it comes with strong assumptions. Most methods assume that we observe all confounders, variables that affect both the causal variables and the outcome variables. But whether we have observed all confounders is a famously untestable assumption. We describe the deconfounder, a way to do causal inference from observational data allowing for unobserved confounding.

How does the deconfounder work? The deconfounder is designed for problems of multiple causal inferences: scientific studies that involve many causes whose effects are simultaneously of interest. The deconfounder uses the correlation among causes as evidence for unobserved confounders, combining unsupervised machine learning and predictive model checking to perform causal inference. We study the theoretical requirements for the deconfounder to provide unbiased causal estimates, along with its limitations and tradeoffs. We demonstrate the deconfounder on real-world data and simulation studies.

02/02/2020

Simon Du presenting on Monday at YINS!

Abstract:
Function approximators, such as deep neural networks, play a crucial role in building intelligent systems that make predictions and decisions. In this talk, I will discuss my work on understanding, designing, and applying function approximators.

First, I will focus on understanding deep neural networks. The main result is that the over-parameterized neural network is equivalent to a new kernel, Neural Tangent Kernel. This equivalence implies two surprising phenomena: 1) the simple algorithm gradient descent provably finds the global optimum of the highly non-convex empirical risk, and 2) the learned neural network generalizes well despite being highly over-parameterized. Furthermore, this equivalence helps us design a new class of function approximators: we transform (fully-connected, graph, convolutional) neural networks to (fully-connected, graph, convolutional) Neural Tangent Kernels, which achieve superior performance on standard benchmarks.

In the second part of the talk, I will focus on applying function approximators to decision-making, aka reinforcement learning, problems. In sharp contrast to the (simpler) supervised prediction problems, solving reinforcement learning problems requires an exponential number of samples, even if one applies function approximators. I will then discuss what additional structures that permit statistically efficient algorithms.

12/09/2019

Yale faculty pitch data science research opportunities.

Statistics & Data Science Project Pitch

Monday, December 9, 2019 | 3:45PM to 5:15PM @ YINS (3rd floor of 17 Hillhouse)

3:45 Refreshments

4:00 Introduction by Daniel Spielman

4:01 - 5:00 Project pitches.

11/18/2019

Moulinath Banerjee from UMich will be presenting tomorrow at Dunham 220!

Information and Abstract:
Single-index type models are popular in statistics, biostatistics and economics as they alleviate the curse of dimensionality to a considerable extent while allowing for broad classes of models through the dependence on an unknown link function. Various classes of single index models are known: in regular models under a fixed dimension setting, the regression parameter is √n estimable; under current status censoring of the response variable in a linear regression model – which leads to the binary choice model — one gets a `generalized’ single index model under heteroscedasticity and a conditional median restriction, with an n^{1/3} convergence rate (Manski’s sccore estimator). Single index structures with a discontinuous link function arise in models involving change-planes in multidimensional space – hyperplanes that separate two (or more) response or survival regimes – and are relevant to applications in personalized medicine and dynamic treatment regimes. Here, rates of estimation can easily exceed √n in the finite dimensional case.

I will talk about some of my recent work in the above class of models in growing and high dimensional settings focusing on how growing dimensions introduce significantly new challenges at both theoretical and computational levels. This is joint work with Debarghya Mukherjee and Ya’acov Ritov.

11/09/2019

Galen Reeves from Duke will be presenting on Monday at Dunham 220!

Information and Abstract:
The analysis of modern high-dimensional inference problems is often framed in terms of detection thresholds, computational-to-statistical gaps, and phase transitions. In this talk, I will describe two scenarios where these properties can be described succinctly using ideas from approximate inference, information theory, and statistical physics. The first part of the talk focuses on random linear estimation problems that arise in channel coding, linear regression, and compressed sensing. I will describe the correspondence between a computationally efficient algorithm (approximate message passing) and the information-theoretic limits. The second part of the talk will focus on the problem of community detection with multiple correlated networks. I will present recent theoretical work showing that the ability to detect different types of community structure can be characterized in terms of the matrix of effective signal-to-noise ratios. I will also discuss some recent methodological approaches inspired by this theory.

11/02/2019

Anru Zhang from UW-Madison will be presenting on Monday at Dunham Lab 220!

Information and Abstract:
The past decade has seen a large body of work on high-dimensional tenors or multiway arrays that arise in numerous applications. In many of these settings, the tensor of interest is high-dimensional in that the ambient dimension is substantially larger than the sample size. Oftentimes, however, the tensor comes with natural low-rank or sparsity structures. How to exploit such structures of tensors poses new statistical and computational challenges.

In this talk, we develop a novel procedure for low-rank tensor regression, namely Importance Sketching Low-rank Estimation for Tensors (ISLET), to address these challenges. The central idea behind ISLET is what we call importance sketching, carefully designed sketches based on both the responses and the structures of the parameter of interest. We show that our estimating method is sharply minimax optimal in terms of the mean-squared error under low-rank Tucker assumptions. In addition, if a tensor is low-rank with group sparsity, our procedure also achieves minimax optimality. Further, we show through numerical studies that ISLET achieves comparable mean-squared error performance to existing state-of-the-art methods whilst having substantial storage and run-time advantages. In particular, our procedure performs reliable tensor estimation with tensors of dimension p = O(10^8) and is 1 or 2 orders of magnitude faster than baseline methods.

10/27/2019

Weijie Sun, University of Pennsylvania, will be presenting tomorrow.

Information and abstract:
Privacy-preserving data analysis has been put on a firm mathematical foundation since the introduction of differential privacy (DP) in 2006, with its deployment on iOS and Chrome lately. This privacy definition, however, has some well-known weaknesses: notably, it does not tightly handle composition. This weakness has inspired several recent relaxations of differential privacy based on Renyi divergences. We propose an alternative relaxation of differential privacy, which we term “f-DP”, which has a number of nice properties and avoids some of the difficulties associated with divergence based relaxations. First, it preserves the hypothesis testing interpretation of differential privacy, which makes its guarantees easily interpretable. It allows for lossless reasoning about composition and post-processing, and notably, a direct way to analyze privacy amplification by subsampling. We define a canonical single-parameter family of definitions within our class that is termed “Gaussian Differential Privacy”, based on hypothesis testing of two shifted normal distributions. We prove that this family is focal to f-DP by introducing a central limit theorem, which shows that the privacy guarantees of any hypothesis-testing based definition of privacy (including differential privacy) converge to Gaussian differential privacy in the limit under composition. This central limit theorem also gives a tractable analysis tool. We demonstrate the use of the tools we develop by giving an improved analysis of the privacy guarantees of noisy stochastic gradient descent. This is joint work with Jinshuo D**g and Aaron Roth.

10/22/2019

Richard Nickl from University of Cambridge will be presenting tomorrow at Dunham Lab 220!

Information and Abstract:
Bayes methods for inverse problems have become very popular in applied mathematics in the last decade after seminal work by Andrew Stuart. They provide reconstruction algorithms as well as in-built “uncertainty quantification” via Bayesian credible sets, and particularly for Gaussian priors can be efficiently implemented by MCMC methodology. For linear inverse problems, they are closely related to classical penalised least squares methods and thus not fundamentally new, but for non-linear and non-convex problems, they give genuinely distinct and computable algorithmic alternatives that cannot be studied by variational analysis or convex optimisation techniques. In this talk we will discuss recent progress in Bayesian non-parametric statistics that allows to give rigorous statistical guarantees for posterior consistency in such models, and illustrate the theory in a variety of concrete non-linear inverse problems arising with partial differential equations.

Address

24 Hillhouse Avenue
New Haven, CT
06511

Opening Hours

Monday	8am - 5pm
Tuesday	8am - 5pm
Wednesday	8am - 5pm
Thursday	8am - 5pm
Friday	8am - 5pm

Telephone

+1 203-432-0666

Website

facebook.com

Alerts

Be the first to know and let us send you an email when Yale Department of Statistics and Data Science posts news and promotions. Your email address will not be used for any other purpose, and you can unsubscribe at any time.

Shortcuts

Want your university to be the top-listed University in New Haven?