Tag Archives: seminars

YINS Tomorrow – 4/18 Sanjeev Arora, Toward theoretical understanding of deep learning

“Toward theoretical understanding of deep learning”

Speaker: Professor Sanjeev Arora

Princeton University & Institute for Advanced Study

Tomorrow – Wednesday, April 18, 2018, 12:00-1:00pm

Location: Yale Institute for Network Science, 17 Hillhouse Avenue, 3rd floor

Abstract: This talk will be a survey of ongoing efforts and recent results to develop better theoretical understanding of deep learning, from expressiveness to optimization to generalization theory. We will see the (limited) success that has been achieved and the open questions it leads to. (My expository articles appear at
http://www.offconvex.org (link is external))

Bio: Sanjeev Arora is Charles C. Fitzmorris Professor of Computer Science at Princeton University and Visiting Professor at the Institute for Advanced Study. He is an expert in theoretical computer science, especially theoretical ML. He has received the Packard Fellowship (1997), Simons Investigator Award (2012), Goedel Prize (2001 and 2010), ACM-Infosys Foundation Award in the Computing Sciences (now called the ACM prize) (2012), and the Fulkerson Prize in Discrete Math (2012).

Upcoming:

4/25/18: Adam Auton (23andme)

5/2/18: Andre Levchenko

Math-applied APPLIED MATH PROGRAM: Seminar & Refreshments Thursday, April 10, 2018

APPLIED MATH/ANALYSIS SEMINAR

Speaker Mauro Maggioni, John Hopkins University

Date: Tuesday, April 10, 2018

Time: 3:45p.m. Refreshments (AKW, 1st Floor Break Area)

4:00p.m. Seminar (LOM 206)

Title: “Learning and Geometry for Stochastic Dynamical Systems in high dimensions”

Abstract:

We discuss geometry-based statistical learning techniques for performing model reduction and modeling of certain classes of stochastic high-dimensional dynamical systems. We consider two complementary settings. In the first one, we are given long
trajectories of a system, e.g. from molecular dynamics, and we estimate, in a robust fashion, an effective number of degrees of freedom of the system, which may vary in the state space of then system, and a local scale where the dynamics is well-approximated by a reduced dynamics with a small number of degrees of freedom. We then use these ideas to produce an approximation to the generator of the system and obtain, via eigenfunctions of an empirical Fokker-Planck equation (constructed from data), reaction coordinates for the system that capture the large time behavior of the dynamics. We present various examples from molecular dynamics illustrating these ideas.

In the second setting we only have access to a (large number of expensive) simulators that can return short paths of the stochastic system, and introduce a statistical learning framework for estimating local approximations to the system, that can be (automatically) pieced together to form a fast global reduced model for the system, called ATLAS. ATLAS is guaranteed to be accurate (in the sense of producing stochastic paths whose distribution is close to that of paths generated by the original system) not only at small time scales, but also at large time scales, under suitable assumptions on the dynamics. We discuss applications to homogenization of rough diffusions in low and high dimensions, as well as relatively simple systems with separations of time scales, and deterministic chaotic systems in high-dimensions, that are well-approximated by stochastic
diffusion-like equations.
Mauro Maggioni 4-10 flyer.pdf

Seminar by Nobel Laureate W.E. Moerner, April 11th

Attached please find a seminar announcement for Nobel Laureate, W.E. Moerner on Wednesday, April 11, 2018.

Speaker: W.E. Moerner, Nobel Laureate

Title: “Single Molecules for 3D Super-Resolution Imaging and Single Particle Tracking in Cells: Methods and Applications:

Date: Wednesday, April 11, 2018

Time & Place: 3:30 PM, SCL 110

Host: Biophysics Training Grant Students

Admins Please Post

Statseminars Joint Biostatistics / Stat & Data Science Seminar , Speaker Carey E. Priebe, 4/9 @4:15pm-5:30pm

biostatistics / STATISTICS & DATA SCIENCE Joint SEMINAR

Date: Monday, April 9, 2018

Time: 4:15pm – 5:30pm

Place: Yale Institute for Network Science, 17 Hillhouse Avenue, 3rd Floor, Rm 328

Seminar Speaker: Carey E. Priebe

Department of Applied Mathematics & Statistics, Johns Hopkins University

Personal Website: https://www.ams.jhu.edu/~priebe/

Title: On Spectral Graph Clustering

Abstract: Clustering is a many-splendored thing. As the ill-defined cousin of classification, in which the observation to be classified X comes with a true but unobserved class label Y, clustering is concerned with coherently grouping observations without any explicit concept of true groupings. Spectral graph clustering — clustering the vertices of a graph based on their spectral embedding — is all the rage, and recent theoretical results provide new understanding of the problem and solutions. In particular, we reset the field of spectral graph clustering, demonstrating that spectral graph clustering should not be thought of as kmeans clustering composed with Laplacian spectral embedding, but rather Gaussian mixture model (GMM) clustering composed with either Laplacian or Adjacency spectral embedding (LSE or ASE); in the context of the stochastic blockmodel (SBM), we use eigenvector CLTs & Chernoff analysis to show that (1) GMM dominates kmeans and (2) neither LSE nor ASE dominates, and we present an LSE vs ASE characterization in terms of affinity vs core-periphery SBMs. Along the way, we describe our recent asymptotic efficiency results, as well as an interesting twist on the eigenvector CLT when the block connectivity probability matrix is not positive semidefinite. (And, time permitting, we will touch on essential results using the matrix two-to-infinity norm.) We conclude with a ‘Two Truths’ LSE vs ASE spectral graph clustering result — necessarily including model selection for both embedding dimension & number of clusters — convincingly illustrated via an exciting new diffusion MRI connectome data set: different embedding methods yield different clustering results, with one (ASE) capturing gray matter/white matter separation and the other (LSE) capturing left hemisphere/right hemisphere characterization.

4:00 p.m. Pre-talk Refreshments

4:15 p.m. – 5:30 Seminar, Room 328, 17 Hillhouse Avenue

For more details and upcoming events visit our website at
http://statistics.yale.edu/ .

Mbbfacultyall FW from Chemistry Dept.: Organic Chemistry Seminar – Daniel Kahne – 3/6/18

On Tuesday, March 6, 2018, Professor Daniel Kahne of Harvard University will be presenting The 2018 Treat B. Johnson Lecture in Organic Chemistry.

The title of his talk is “Molecular Machines That Build Membranes”.

You are welcomed to attend the lecture which will be held at The Sterling Chemistry Lab – 225 Prospect Street – in SCL 110.

If you cannot attend, you can watch by going to the following link:

https://yale.hosted.panopto.com/Panopto/Pages/Viewer.aspx?id=848c083b-b7cc-4b48-93ed-a897010fbd2e
KAHNE Poster.pdf

NIMH Virtual Workshop on Quantum Computing

NIMH VIRTUAL WORKSHOP:

SOLVING COMPUTATIONAL CHALLENGES IN GENOMICS AND NEUROSCIENCE VIA PARALLEL & QUANTUM COMPUTING

March 28, 2018

9:00 am – 1:00 pm EST

Goal of the workshop

This virtual workshop aims to highlight core computational problems faced by genetics and the subdomains of neuroscience that parallel or quantum computing can address. By bringing together experts in quantum and parallel computing with experts in genetics and neuroscience, we hope to start a dialogue between academic and industry partners working in this area with the focus on algorithm optimization and development. This virtual workshop will be the forum and the nexus to find convergence between cross-disciplinary fields that are operating mostly independently – 1) genomics and neuroscience, and 2) AI/machine learning and 3) quantum computing. The goal is to identify key avenues for computation optimization via parallel and quantum algorithms. This workshop will facilitate the use of state-of-art computational technologies for addressing core bottlenecks in genomics and neuroscience.

Overview

This workshop will cover the following topics with 5 minutes break following each topic discussion:

  • Opening Remarks (10 min)
  • Topic 1: Computational Challenges in Genetics and Neuroscience (1.5 hour)
  • Topic 2: AI, machine learning and parallel computing (45 min)
  • Topic 3: Quantum Algorithms for Accelerated Computation: Opportunities and Challenges (1 hour)
  • Roundtable Discussion & Summary (30 mins)

*NOTE: Some speakers are yet to be confirmed and/or subject to change.

9:00 – 9:10 am:Opening Remarks – Thomas Lehner, Geetha Senthil, Susan

Wright, National Institute of Mental Health, Office of Genomics Research Coordination

Morning Session

Chairs: Alan Anticevic, Ph.D., Yale University and Alan Aspuru-Guzik, Ph.D., Harvard University

Topic 1: Computational Challenges in Genetics and Neuroscience

This session is to highlight where computational challenges/bottlenecks exist at the level of scaling (data and computational features) and computational speedup.

9:10 – 9:25 am: Presentation 1: Genetics and functional genomics

Michael McConnell, Ph.D., University of Virginia, Michael Gandal, M.D., Ph.D., University of California, Los Angeles

9:25 – 9:40 am: Presentation 2: Neurophysiology (processing data, extracting, analysis)

Potential speakers: Mike Halassa, M.D., Ph.D., Massachusetts Institute of Technology

9:40 – 9:55 am: Presentation 3: Neuroimaging

Potential speakers: Alan Anticevic, Ph.D., Yale University, Stephen Smith, Oxford

9:55 – 10:10 am: Presentation 4: Quantitative deep phenotypic analysis

Potential speakers: Andrey Rzhetsky, Ph.D., University of Chicago, Justin Baker, M.D., Ph.D., Massachusetts General Hospital, Jukka-Pekka Onnela, M.Sc., Ph.D., Harvard University

10:10 – 10:25 am: Presentation 5: Computational modeling

Suggested topic: Spiking and neural models and ion channel modelling – spiking network simulation

Speakers: John Murray, Ph.D., Yale University, Michael Hines, Ph.D., Yale University

10:25 – 10:30 am: Break

Topic 2: AI, machine deep learning and parallel computing

This session is to discuss application of state-of-the-art classical parallel computing algorithm applications for machine learning, simulation, & optimization of analysis with ‘big’ data.

10:30 – 10:45 am: Presentation 1: Overview of machine learning via classical and parallel computing technologies

Potential speakers: Guillermo Sapiro, M.Sc., Ph.D., Duke University

10:45 – 11:00 am: Presentation 2: Deep Learning for AI applications – e.g. DeepMind

Potential speakers: Tim Lillicrap, Ph.D., DeepMind

11:00 – 11:15 am: Presentation 3: Parallel processing & GPUs

Suggested topic: Nvidia parallel processing & GPU capabilities for efficient high-performance applications

Potential speakers: Alan will reach out to his contact at Nvidia

11:15 – 11:20 am: Break

Afternoon Session:

Chairs: Aram Harrow, Ph.D., Massachusetts Institute of Technology, and John Murray, Ph.D.,

Yale University

Topic 3: Quantum Algorithms for Accelerated Computation: Opportunities and Challenges

This session will discuss the current state of quantum hardware and algorithms. What kind of advantages (either in terms of speed or solution quality) can be obtained by using quantum machine learning? How close are existing or proposed near-term hardware platforms to being able to implement these algorithms?

11:20 – 11:35 am: Presentation 1: Overview and primer: what is quantum computing good for?

Potential speakers: Alán Aspuru-Guzik, Ph.D., Harvard University

11:35 – 11:50 am: Presentation 2: Status and Prospects for Quantum Hardware

Potential speaker: Nicole Barberis, IBM

11:50 am – 12:05 pm: Presentation 3: Promising Quantum Computing Algorithms on the

Horizon

Potential speakers: Ashley Montanaro, Ph.D., University of Bristol

12:05 – 12:20 pm: Presentation 4: Quantum Machine Learning and Optimization

Seth Lloyd, Massachusetts Institute of Technology

12:20 – 12:30 pm: Break

12:30 – 12:50 pm: Roundtable Discussion & Summary

Moderators: Stefan Bekiranov, University of Virginia & John Murray, Yale University

  • What are the immediate avenues for computation optimization via parallel computing?
  • Which problems are suitable for parallel vs. quantum computing?
  • What are the distinct challenges facing parallel vs quantum computing platforms?
  • Which are the most impactful avenues for quantum algorithm development from the standpoint of neuroscience and genomics?
  • Opportunities for public private partnership?

    12:50 – 1:00 pm: Summary/Closing Remarks

    Potential speakers: Alán Aspuru-Guzik, Harvard University, Alan Anticevic, Yale University

    1:00 pm: Adjourn

NIMH Quantum Computing Virtual Workshop Agenda_02-15-2018.docx

Secondary_appt Department-cs CS Colloquium/Danqi Chen, Stanford Univ./Feb. 26, 4pm/AKW 200

CS Colloquium

Monday, February 26

4:00 p.m., AKW 200 (coffee & cookies at 3:45)

Speaker: Danqi Chen, Stanford University

Title: Knowledge from Deep Understanding of Language

Host: Dragomir Radev

Abstract:

Almost all of humanity’s knowledge is now available online, but the vast majority of it is principally encoded in the form of human language explanations. In this talk, I explore novel neural network or deep learning approaches that open up increased opportunities for getting a deep understanding of natural language text. First, I show how distributed representations enabled the building of a smaller, faster, better dependency parser for finding the structure of human language sentences. Then I show how related neural technologies can be used to improve the construction of knowledge bases from text. However, maybe we don’t need this intermediate step and can directly gain knowledge and answer people’s questions from large textbases? In the third part, I explore doing this by looking at a simple but highly effective neural architecture for question answering.

Bio:

Danqi Chen is a PhD student in Computer Science at Stanford
University, working with Christopher Manning on deep learning approaches to Natural Language Processing. Her research centers on how computers can achieve a deep understanding of human language and the information it contains. Danqi received Outstanding Paper Awards at ACL 2016 and EMNLP 2017, a Facebook Fellowship, a Microsoft Research Women’s Fellowship and an Outstanding Course Assistant Award from Stanford. She holds a B.E. with honors from Tsinghua University.