Upcoming ISE Seminars

Please take the opportunity to visit the ISE Department, have a drink with your fellow classmates and professors and meet graduating students! RSVP to aeb213@lehigh.edu if you are planning to stop by!

We present a model for clustering which combines two criteria: Given a collection of objects with pairwise similarity measure, the problem is to find a cluster that is as dissimilar as possible from the complement, while having as much similarity as possible within the cluster. The two objectives are combined either as a ratio or with linear weights. The ratio problem, and its linear weighted version, are solved by a combinatorial algorithm within the complexity of a single minimum s,t-cut algorithm. This problem (HNC) is closely related to the NP-hard problem of normalized cut that is often used in image segmentation and for which heuristic solutions are generated with the eigenvector technique (spectral method).

It is demonstrated that HNC is a form of relaxation of normalized cut and its generalization to "q-normalized cut". We study the relationship between this HNC relaxation and the spectral method and demonstrate a number of advantages for the combinatorial algorithm. These advantages include a better approximation, in practice, of the normalized cut objective for image segmentation benchmark problems.

HNC can be utilized as a supervised or unsupervised machine learning technique. It has been used for data mining, and its comparison to leading machine learning techniques on datasets selected from the UCI data mining benchmark and other benchmarks indicates that its use of pairwise comparisons is powerful in improving performance accuracy. HNC is currently the leading neuron segmentation algorithm in the Neurofinder benchmark for cell identification in calcium imaging movies. Time permitting, we will discuss our recently developed methods employed to make the family of HNC algorithms scalable for massive scale data sets.

Please join the ISE Department at the annual ISE Banquet! This year's cocktail reception & banquet will be held in the ASA Packer Dining Room, University Center. The Cocktail Reception will begin at 5:30 pm and then the banquet at 6:30 pm!

If you are interested in attending, please register here: https://www.eventville.com/catalog/eventregistration1.asp?eventid=1012138.

MOSEK is a software package for solving large scale sparse optimization problems. To be precise MOSEK is capable of solving linear, convex quadratic and conic quadratic optimization problems possibly having some integer constrained variables. In addition MOSEK can solve continuous semidefinite optimization problems.

In this presentation we will review what is new and improved in the recently released version 8. We will also present computational results that documents upgrading to MOSEK version 8 provides a genuine enhancement of the numerical stability and performance.

The results also documents that MOSEK is one of the best if not the best semidefinite optimizer.

Product and content personalization is now ubiquitous in e-commerce. Available transactional data is typically too sparse for this task. As such, companies today seek to use a variety of information on the interactions between a product and a customer to drive personalization decisions. We formalize this problem as one of recovering a large-scale matrix, with side information in the form of additional matrices of conforming dimension. Viewing the matrix we seek to recover and the side information we have as slices of a tensor, we consider the problem of Slice Recovery, which is to recover specific slices of ‘simple’ tensors from noisy observations of the entire tensor. We propose a definition of simplicity that on the one hand elegantly generalizes a standard generative model for our motivating problem, and on the other subsumes low-rank tensors for a variety of existing definitions of tensor rank. We provide an efficient algorithm for slice recovery that is practical for massive datasets and provides a significant performance improvement over state of the art incumbent approaches to tensor recovery. Further, we establish near-optimal recovery guarantees that in an important regime represent an order improvement over the best available results for this problem. Experiments on data from a music streaming service demonstrate the performance and scalability of our algorithm.

Lehigh’s Industrial and Systems Engineering (ISE) Council will be holding their seventh ISE Career Fair on September 14, 2016, a day before the Lehigh University Career Fair on September 15th. Employers and students will be able to meet in a personal setting that and discuss the company’s internship/co-op/job opportunities! *This event is for both ISE & HSE (Healthcare Systems Engineering) students. All companies attending will receive a resume book. Registration fees are per company. Sponsor - $130 Sponsors of the Networking event will get a 5-minute presentation during the first half hour of the event. Sponsorship is limited! Regular - $80, Government - $50, and Non-Profit - $30 (must send a copy of 501 (c) 3).

Register here: https://www.eventville.com/catalog/eventregistration1.asp?eventid=1011949

Thursday, April 28, 2016 - 5:00pm
Please join the ISE Department at the annual ISE Banquet.
A Cocktail Reception will begin at 5:00 p.m. in the Iacocca Hall, Siegel Gallery (Ground Floor) followed by the Banquet Dinner at 6:00 p.m. in Wood Dining Room, Iaccocca Hall.
If you are interested in attending, please register here: https://www.eventville.com/catalog/eventregistration1.asp?eventid=1011889

This talk is an “eye witness account” of the evolution of nonlinear optimization methods over the last 4 decades. Starting from the early days of simplex-inspired methods, to augmented Lagrangians and interior-points, this talk highlights the need for new active set methods and the opportunities (not challenges) provided by stochasticity. We conclude with a few observations about the formulation and solution of optimization problems in supervised learning.

In this talk I will show some very recent results on optimal strategies for betting on individual sequences of binary outcomes, that is betting against a non-stochastic coin. This naturally extends the well-known Kelly strategy to the adversarial domain.
Moreover, I will show some surprising links between betting, online learning, and adaptive stochastic optimization. Solving optimally coin betting will allow to solve optimally all these problems, with the very same algorithm.

Empirically results will be shown as well.