Seminars

Studentprojekten för Självständigt arbete och Uppsatsmetodik på IT-programmets årskurs 3 har posterutställning i korridoren på plan 2 i hus 2. Varmt välkomna!

Language: English

Abstract
Estimation of the parameters of chirp signal model is a fundamental problem in signal processing and has been of interest for a long time. Chirp signals embedded in additive noise occur in a variety of signal processing applications, e.g. in SAR, inverse SAR imaging, sonar and communications. Given a set of observed signals, the problem is to estimate the unknown frequencies and the frequency rates associated with such a chirp model. In this talk, we present some recent results in this area. Theoretical asymptotic properties of important estimators of the parameters of a real valued non-linear chirp signal model will be discussed.
We first consider the non-linear least squares estimators (NLSE) of the parameters of a chirp model in the presence of a stationary additive noise. Using a number theoretic result, we present a simplified form of the asymptotic variance covariance matrix of the NLSE. We also present a sequential estimation procedure for obtaining the NLSE of a superimposed chirp model and establish, asymptotically, that the proposed estimators are strongly consistent.
We further study the asymptotic theoretical properties of robust least absolute deviation (LAD) estimators of the parameters of a chirp model in the presence of additive white noise. We establish that the LAD estimators are strongly consistent, find the convergence rates and present the form of the asymptotic joint distribution of the resultant estimators.
A computationally efficient algorithm for estimating the frequencies and the frequency rates of a chirp signal model in the presence of stationary noise will also be presented. We show that the proposed iterative algorithm yields efficient estimators in a fixed number of iterations. Strong consistency results and asymptotic equivalence of the proposed estimators and the NLSE of the corresponding parameters will be presented.

CV

All are welcome!

Abstract
Opponent: Professor Anne Persson, Forskningscentrum för Informationsteknologi, Högskolan i Skövde.

Abstract:
GeoMemories is an online application for visualizing the geospatial changes of the Italian landscape. Recently an article was accepted for a journal, which shows the challenges of stitching millions of photos together. It also give examples of how the image archive, hosted by the Aerofototeca Nazionale in Rome together with other photo archives, can be used to monitor the long time change of the landscape. Some improvements regarding image stitching will also be explained.

Opponent: Dr. Xing Cai, Senior Research Scientist at Simula Research Laboratory and Professor at the University of Oslo

The discussion will be held in English.

Abstract: Modern computer architectures, with multicore CPUs and GPUs or other accelerators, make stronger demands than ever on writers of scientific code. Normally, the most efficient program has to be written - using a substantial effort - by expert programmers for a certain application on a particular computer. This thesis deals with several algorithmic and technical approaches towards effectively satisfying the demand for high performance parallel scientific applications on hybrid computer architectures without incurring such a high cost in expert programmer time. Efficient programming is accomplished by writing performance-portable code where performance-critical functionality is provided either by an optimized library or by adaptively selecting which computational tasks that are executed on the CPU and the accelerator.

Language: English

Abstract
Traditional life data analysis analyzes time-to-failure data (of a product, system or component) obtained under normal operating conditions to quantify the life characteristics of the objects. Accelerated life test (ALT) is a popular experimental strategy to obtain information on life distributions of highly reliable products. The main idea is to submit materials to higher than usual environmental conditions, or stress, to ensure early failure. Data obtained from such an experiment need to be extrapolated to estimate lifetime distribution under normal conditions.
A Step-Stress life test is a particular type of ALT. We observe the failure times of the products at a particular stress level, and then change the stress to a different level. Failure times in the new stress level are observed, and the stress level is changed again and so on. There are several stress loading schemes - constant stress, step-stress, ramp-stress, progressive stress, etc. Introduction of stresses ensures a reasonable number of failures and reduces the experimental time. The stress structure and the lifetime distribution have to be combined to get the comprehensive ALT model.
The analysis relies on life and stress data or time-to-failure data at a specific stress level. Data are either complete or censored. In this talk we discuss various aspects of the exact inference for the two-parameter Exponential distribution (a standard life-time distribution) under Type-II Hybrid Censoring (a popular censoring method) and Bayesian analysis of different Hybrid and Progressive Life Tests. This would be followed by exact and approximate inference procedures for a Simple Step-Stress model under Exponential distribution with location and scale parameters. We shall further consider the order restricted Bayesian inference for the Exponential Simple Step-Stress Model. Finally, Bayesian analysis of Simple Step-stress Model under Weibull lifetimes (another standard lifetime distribution) would be discussed in brief.

CV

All are welcome!

Abstract

Abstract
Opponent: Doctor Koichi Takahashi, RIKEN, Quantitative Biology Center.

Abstract