See all upcoming seminars in LäsIT and seminar web pages at the homepage for the PhD studentseminars, TDB, Vi2, Theory and Applications Seminars (TAS) @ UpMARC., Department of Mathematics and The Stockholm Logic Seminar.
Tomorrow (22 Nov)
|Carl Nettelblad: Hidden Markov Models for Genotype Phasing and Handwritten Text Alignment|
Location: ITC 2446, Time: 13:15-14:00
Abstract: While there are more recent forms of statistical models for complex stochastic processes, hidden Markov models are still useful in many fields. I will discuss two applications, for analyzing the transmission of genetic material between parents and offspring, and a nascent collaboration with Vi2 regarding using a hidden Markov model as a "proofreading" stage when identifying words in written text. In addition, I will mention why you would want to solve a small ODE in a modification of the expectation-maximization (EM) parameter estimation algorithm.
Thursday 23 Nov
|Ghafour Ahani: Cost-Optimal Caching for D2D Networks with Presence of User Mobility|
Location: ITC 1245, Time: 13:00-14:00
Abstract: Nowadays, there is a heavy burden on the backhaul networks due to exponential data traffic growth. A promising approach is caching files at the user equipments (e.g., mobile phones) due to massive device connectivity. Users can download their files from each other through device-to-device communications. In this seminar, an optimal caching problem with respect to user mobility will be discussed.
Friday 24 Nov
|Bengt Jonsson: Model Learning: Generating Automata Models from Tests|
Location: ITC 1245, Time: 14:15
Abstract: Model-based approaches to development, verification, and testing are becoming increasingly important for efficient development of reliable software Its application is hampered by a lack of adequate specifications for software components, libraries, and services. This problem is addressed by the area of Model Learning, also known under names such as "Specification Mining" or "Test-Based Modeling". More technically, and in this context, Model Learning consists of techniques for generating automata models from outcomes of tests on a black-box component. This presentation will review some basic principles of model learning, and present an overview of recent results and work in progress by the Uppsala team (also in collaboration with the groups of Bernhard Steffen (TU Dortmund) and Frits Vaandrager (U Nijmegen)). We present how results on learning finite-state models can be extended to the learning of infinite-state models, e.g., to capture the influence of data value on the dynamic behavior of a component, or the influence of timers. We also survey some recent applications of this generalization to learning models of, communication protocols and library components.
Monday 27 Nov
|Jean-Noël Monette, (Tacton Systems AB): CP for Product Configuration at Tacton|
Location: ITC 1311, Time: 10:15
Abstract: Tacton is a world leader in product configuration, where a configuration engine is used to configure a complex product (e.g., a truck) in order to meet the users requirements (e.g., the truck should be used mostly on highways) and respect the established rules (e.g., compatibility between wheels and suspensions). A configuration engine helps the user by guiding him through the potentially huge space of feasible configurations (e.g, by presenting a default solution, by marking which values can be selected, or by resolving potential conflicts). To do so, the Tacton configuration engine maps configuration problems to constraint satisfaction problems (CSP) and uses an off-the-shelf constraint programming solver.
In this presentation, I will first present Tacton, what product configuration is, and how we help businesses be more effective. I will then describe and contrast existing approaches to product configuration, before focusing on Tacton's constraint-based solution.
Monday 27 Nov
|Axel Ringh: Optimal Mass Transport as a Distance Measure between Images|
Location: ITC 4307, Time: 14:15
Abstract: The optimal mass transport problem is a geometric framework for how to transport masses in an optimal way. Historically it has had large impact in economic theory and operations research, and recently it has also gained significant interest in application areas such as signal processing, image processing, and machine learning. The optimal mass transport problem can be formulated as a linear programming problem, however when computing the distance between two images the size of this linear program becomes prohibitively large. A recently development to address this builds on using an entropic barrier term and solving the resulting optimization problem using so called Sinkhorn iterations. This allows for an approximate solution of large optimal mass transport problems. In this work we show how these results can be used and extended in order to use optimal mass transport for solving inverse problems in, e.g., computerized tomography.
Monday 27 Nov
| Olle Terenius: Disseminate your research with Wikipedia|
Location: ITC 4307, Time: 15:15
|Gender Equality Group - Monthly Meeting|
| Location: ITC 1345, Time: 13:30-14:30|
ON THE AGENDA:
See also the list of all upcoming seminars.
Internal seminars. Lecturers may be either internal or external.