Loading Events

« All Events

  • This event has passed.

Autumn School: Optimization in Machine Learning and Data Science

August 28 - August 31



Conference Fee: 70€

autumn school graphicLarge parts of the new economy rely on data science and the application of machine learning. For handling these modern computational and statistical problems, optimization algorithms are of utmost importance. The typically large data sets involved in this context urge for new optimization techniques.

Therefore, we are delighted to announce Julian Hall,  Tamara G. Kolda, Stephen Wright as speakers at the upcoming ALOP Autumn School entitled “Optimization in Machine Learning and Data Science.”

Representing three top class authorities in their respective fields of research, reaching from optimization of signal and image processing over large scale linear programming problems to tensor decomposition and many more, they will be sharing their experience and knowledge on theory and application of optimization algorithms.

Autumn School 2017 Program (click for download)

Registration is closed

Information on scheduled Speakers: 

Julian_teachingJulian Hall, The University of Edinburgh – 

His research interests are the development of algorithmic and computational techniques for solving large scale linear programming (LP) problems using the revised simplex method on both serial and parallel computers. Consequential research interest is the application of these techniques in other areas of computation optimization and linear algebra.

Mr. Hall plans on talking about high performance simplex which will naturally lead into high performance numerical computation. He will look at computational issues in interior point methods and the applicability of “modern” first order methods to the solution of LP problems, as well as talk about one of his current research topics.

 

 Tamara G. Kolda, Sandia National Laboratories – 

Her research interests are computational algorithm design and development, including linear and multi-linear algebra, tensor decomposition, tensor eigenvalues, graph algorithms, machine learning, network science, derivative-free optimization, computational optimization, distributed and parallel computing.Tammy_teaching

Ms. Kolda will talk about “Optimization Approaches for Fitting the Canonical Tensor Decomposition.” More specifically, she will speak about

  • the theory and background of the CP decomposition
  • standard approaches such as alternating least squares and all-at-once optimization
  • handling missing data
  • practical approaches to choosing the rank
  • alternative objective functions for other statistical models/assumptions
  • Time permitting, she may also talk about HPC, data structures, parallelization

 

Stephen J. Wright, University of Wisconsin-Madison

Steve_teaching

His research interests are in continuous optimization algorithms and their applications. He works on devising algorithms for fundamental problems in continuous optimization (with practical relevance) and in understanding their theoretical properties. He also enjoys working with domain scientists and engineers in such areas as data analysis and machine learning, machine learning, process control, and computer architecture to solve problems in these areas using formulation and solution techniques from optimization.

Mr. Wright will cover the background of fundamental algorithmic approaches that are relevant to data analysis problems. Topics will include the following:

– first-order (gradient) methods for smooth unconstrained   minimization.
   Steepest descent, accelerated gradient and other momentum methods.

– minimization of objectives with “simple” regularization terms using
   forward-backward/prox-linear methods.

– stochastic gradient and subgradient methods for minimizing convex objectives with “summation” form.

– second-order methods for smooth (convex and nonconvex) objectives.  Newton’s method and methods that explicitly use
  negative curvature information.

We will also give a survey of current use of optimization methodology in data analysis.

 

Organizers: 

Jan Pablo Burgard, Ralf Münnich, Ekkehard Sachs, Volker Schulz, Sven de Vries 

Schedule, Location and Travel Information: 

Feel free to download a copy of the schedule here:  Autumn School 2017 Program

The workshop is scheduled to begin on Monday, August 28 at 9:00 am and end on Thursday, August 31 in the afternoon.
The workshop will take place in Building E located on Campus I at Trier University.
For travel tips on how to get to Trier, please check  here;
Instructions on how to find Campus I are located here.

 

Hotel Room Reservations

We have a reserved a hotel room contingent at the Altstadt Hotel, Am Porta Nigra Platz, 54292 Trier. Please contact them directly to make your reservations referencing the AUTUMN SCHOOL 2017.

The contingent will be available for reservation until August 1, 2017, or until all available rooms have been booked.

Registration information:

There will be a registration fee of 70 €. Detailed payment information will be communicated upon registration.

Qualified participants may be eligible for a stipend to assist with travel expenses. In order to be considered for the stipend, please submit a CV (which should include the names and contact information of two professional references) along with a one-page letter of motivation before June 15, 2017.

Registration is closed

Details

Start:
August 28
End:
August 31
Cost:
70€
Event Category:

Organizer

Research Training Group ALOP at Trier University
Phone:
0651-2013461
Email:
alop@uni-trier.de

Venue

Trier University, E Building
Universitätsring 15
Trier, Germany
+ Google Map


ALOP