Tommi Kärkkäinen

Guided ultrasonic waves in long bones: modelling, experiment and in vivo application

Existing ultrasound devices for assessing the human tibia are based on detecting the first arriving signal, corresponding to a wave propagating at, or close to, the bulk longitudinal velocity in bone. However, human long bones are effectively irregular hollow tubes and should theoretically support the propagation of more complex guided modes similar to Lamb waves in plates. Guided waves are attractive because they propagate throughout the bone thickness and can potentially yield more information on bone material properties and architecture. In this study, Lamb wave theory and numerical simulations of wave propagation were used to gain insights into the expected behaviour of guided waves in bone. Experimental measurements in acrylic plates, using a prototype low-frequency axial pulse transmission device, confirmed the presence of two distinct propagating waves: the first arriving wave propagating at, or close to, the longitudinal velocity, and a slower second wave whose behaviour was consistent with the lowest order Lamb antisymmetrical (A0) mode. In a pilot study of healthy and osteoporotic subjects, the velocity of the second wave differed significantly between the two groups, whereas the first arriving wave velocity did not, suggesting the former to be a more sensitive indicator of osteoporosis. We conclude that guided wave measurements may offer an enhanced approach to the ultrasonic characterization of long bones.

An enhanced memetic differential evolution in filter design for defect detection in paper production

This article proposes an Enhanced Memetic Differential Evolution (EMDE) for designing digital filters which aim at detecting defects of the paper produced during an industrial process. Defect detection is handled by means of two Gabor filters and their design is performed by the EMDE. The EMDE is a novel adaptive evolutionary algorithm which combines the powerful explorative features of Differential Evolution with the exploitative features of three local search algorithms employing different pivot rules and neighborhood generating functions. These local search algorithms are the Hooke Jeeves Algorithm, a Stochastic Local Search, and Simulated Annealing. The local search algorithms are adaptively coordinated by means of a control parameter that measures fitness distribution among individuals of the population and a novel probabilistic scheme. Numerical results confirm that Differential Evolution is an efficient evolutionary framework for the image processing problem under investigation and show that the EMDE performs well. As a matter of fact, the application of the EMDE leads to a design of an efficiently tailored filter. A comparison with various popular metaheuristics proves the effectiveness of the EMDE in terms of convergence speed, stagnation prevention, and capability in detecting solutions having high performance.

A Memetic Differential Evolution in Filter Design for Defect Detection in Paper Production

This article proposes a Memetic Differential Evolution (MDE) for designing digital filters which aim at detecting defects of the paper produced during an industrial process. The MDE is an adaptive evolutionary algorithm which combines the powerful explorative features of Differential Evolution (DE) with the exploitative features of two local searchers. The local searchers are adaptively activated by means of a novel control parameter which measures fitness diversity within the population. Numerical results show that the DE framework is efficient for the class of problems under study and employment of exploitative local searchers is helpful in supporting the DE explorative mechanism in avoiding stagnation and thus detecting solutions having a high performance.

Fitness diversity based adaptation in Multimeme Algorithms:A comparative study

This paper compares three different fitness diversity adaptations in multimeme algorithms (MmAs). These diversity indexes have been integrated within a MmA present in literature, namely fast adaptive memetic algorithm. Numerical results show that it is not possible to establish a superiority of one of these adaptive schemes over the others and choice of a proper adaptation must be made by considering features of the problem under study. More specifically, one of these adaptations outperforms the others in the presence of plateaus or limited range of variability in fitness values, another adaptation is more proper for landscapes having distant and strong basins of attraction, the third one, in spite of its mediocre average performance can occasionally lead to excellent results.

Automated software license analysis

Software license is a legal instrument governing the usage or redistribution of copyright-protected software. License analysis is an elaborate undertaking, especially in case of large software consisting of numerous modules under different licenses. This paper describes an automated approach for supporting software license analysis. The approach is implemented in a reverse engineering tool called ASLA. We provide a detailed description of the architecture and features of the tool. The tool is evaluated on the basis of an analysis of 12 OSS (open source software) packages. The results show that licenses for (on average) 89% of the source code files can be identified by using ASLA and that the efficiency of the automated analysis is (on average) 111 files per second. In a further comparison with two other open source license analyzers—OSLC and FOSSology—ASLA shows a competitive performance. The results validate the general feasibility of the ASLA approach in the context of analyzing non-trivial OSS packages.

Independent component analysis on the mismatch negativity in an uninterrupted sound paradigm

We compared the efficiency of the independent component analysis (ICA) decomposition procedure against the difference wave (DW) and optimal digital filtering (ODF) procedures in the analysis of the mismatch negativity (MMN). The comparison was made in a group of 54 children aged 8-16 years. The MMN was elicited in a passive oddball protocol presenting uninterrupted auditory stimulation consisting of two frequent alternating tones (600 and 800 Hz) of 100 ms duration each. Infrequently, one of the 600 Hz tones was shortened to 50 or 30 ms. The event related potentials (ERPs) were decomposed into the MMN-like and non-MMN-like independent components (ICs) through the FastICA algorithm. The ICA decomposition procedure extracted a cleaner MMN compared to the ODF or DW procedures. It extracted the MMN, whose characteristics concurred with the substantial number of publications demonstrating a significantly larger peak amplitude and shorter latency of the MMN in response to the more deviant stimulus (30 ms) compared to the less deviant stimulus (50 ms). The MMN to these two deviant stimuli did not differ in the peak amplitude or latency when it was extracted through the other two procedures. The ICA decomposition and ODF procedures, similarly, significantly improved the single trial signal-to-noise ratio (SNR) of the MMN compared to the DW procedure. Due to this improvement, the proposed ICA decomposition procedure might allow shortening of the recording session and could be used to study the MMN in paradigms similar to this with small modifications.

Robust Formulations for Training Multilayer Perceptrons

The connection between robust statistical estimates and nonsmooth optimization is established. Based on the resulting family of optimization problems, robust learning problem formulations with regularization-based control on the model complexity of the multilayer perceptron network are described and analyzed. Numerical experiments for simulated regression problems are conducted, and new strategies for determining the regularization coefficient are proposed and evaluated.

Denoising of Smooth Images Using L 1 -Fitting

In this paper, denoising of smooth (H10-regular) images is considered. The purpose of the paper is basically twofold. First, to compare the denoising methods based on L1- and L2-fitting. Second, to analyze and realize an active-set method for solving the non-smooth optimization problem arising from the former approach. More precisely, we formulate the algorithm, proof its convergence, and give an efficient numerical realization. Several numerical experiments are presented, where the convergence of the proposed active-set algorithm is studied and the denoising properties of the methods based on L1- and L2-fitting are compared. Also a heuristic method for determining the regularization parameter is presented and tested.

Comparison of formulations and solution methods for image restoration problems

Different formulations for image restoration problems are presented, analysed and compared. Two of the three formulations considered are smooth enough to satisfy the assumptions for convergence of ordinary gradient-based optimization methods, such as the conjugate gradient method. For solving the third problem, two general methods of nonsmooth optimization are applied: a (first-order) proximal bundle method and a (second-order) bundle-Newton method. Moreover, a new generalization of active-set methods that have earlier shown high efficiency for BV-regularized image restoration problems is proposed and analysed. Comparison of methods as well as different formulations is completed with numerical experiments.

Online mass flow prediction in CFB boilers with explicit detection of sudden concept drift

Fuel feeding and inhomogeneity of fuel typically cause fluctuations in the circulating fluidized bed (CFB) process. If control systems fail to compensate the fluctuations, the whole plant will suffer from dynamics that is reinforced by the closed-loop controls. This phenomenon causes reducing efficiency and the lifetime of process components. In this paper we address the problem of online mass flow prediction, which is a part of control. Particularly, we consider the problem of learning an accurate predictor with explicit detection of abrupt concept drift and noise handling mechanisms. We emphasize the importance of having domain knowledge concerning the considered case and constructing the ground truth for facilitating the quantitative evaluation of different approaches. We demonstrate the performance of change detection methods and show their effect on the accuracy of the online mass flow prediction with real datasets collected from the experimental laboratory-scale CFB boiler.