Multi-objective performance optimisation for model predictive control by goal attainment

This article proposes an approach for performance tuning of model predictive control (MPC) using goal-attainment optimisation of the cost function weighting matrices. The approach is developed for three formulations of the control problem: (i) minimal and (ii) non-minimal design based on the same cost function and (iii) a non-minimal MPC approach with an explicit integral-of-error state variable and modified cost function. This approach is based on earlier research into multi-objective optimisation for proportional-integral-plus control systems. Simulation experiments for a 3-input, 3-output Shell heavy oil fractionator model illustrate the feasibility of MPC goal attainment for multivariable decoupling and attainment of a specific output response. For this example, the integral-of-error state variable offers improved design flexibility and hence, when it is combined with the proposed tuning method, yields an improved closed-loop response in comparison to minimal MPC.     

Parsing human skeletons in an operating room

Multiple human pose estimation is an important yet challenging problem. In an operating room (OR) environment, the 3D body poses of surgeons and medical staff can provide important clues for surgical workflow analysis. For that purpose, we propose an algorithm for localizing and recovering body poses of multiple human in an OR environment under a multi-camera setup. Our model builds on 3D Pictorial Structures and 2D body part localization across all camera views, using convolutional neural networks (ConvNets). To evaluate our algorithm, we introduce a dataset captured in a real OR environment. Our dataset is unique, challenging and publicly available with annotated ground truths. Our proposed algorithm yields to promising pose estimation results on this dataset.     

Optimizations of the naïve-Bayes classifier for the prognosis of B-Chronic Lymphocytic Leukemia incorporating flow cytometry data

Prognosis of B-Chronic Lymphocytic Leukemia (B-CLL) remains a challenging problem in medical research and practice. While the parameters obtained by flow cytometry analysis form the basis of the diagnosis of the disease, the question whether these parameters offer additional prognostic information still remains open. In this work, we attempt to provide computer-assisted support to the clinical experts of the field, by deploying a classification system for B-CLL multiparametric prognosis that combines various heterogeneous (clinical, laboratory and flow cytometry) parameters associated with the disease. For this purpose, we employ the na?ve-Bayes classifier and propose an algorithm that improves its performance. The algorithm discretizes the continuous classification attributes (candidate prognostic parameters) and selects the most useful subset of them to optimize the classification accuracy. Thus, in addition to the high classification accuracy achieved, the proposed approach also suggests the most informative parameters for the prognosis. The experimental results demonstrate that the inclusion of flow cytometry parameters in our system improves prognosis.     

The SAINT observatory subsystem: an open-source intelligence tool for uncovering cybersecurity threats

International Journal of Information Security - Data from Online Social Networks, search engines, and the World Wide Web are forms of unstructured knowledge that are not regularly used in...     

Time series analysis with multiple resolutions

We introduce a new representation for time series, the Multiresolution Vector Quantized (MVQ) approximation, along with a distance function. Similar to Discrete Wavelet Transform, MVQ keeps both local and global information about the data. However, instead of keeping low-level time series values, it maintains high-level feature information (key subsequences), facilitating the introduction of more meaningful similarity measures. The method is fast and scales linearly with the database size and dimensionality. Contrary to previous methods, the vast majority of which use the Euclidean distance, MVQ uses a multiresolution/hierarchical distance function. In our experiments, the proposed technique consistently outperforms the other major methods.     

A Multiagent Platform for Ubiquitous Service Provision

The Virtual Home Environment is very important in contemporary mobile telecommunications infrastructure as it caters for the ubiquitous provision of services irrespective of network, location and user device. The universality of systems like Universal Mobile Telecommunications System and wi-fi increases the need for the rapid introduction of efficient VHE schemes. In this paper, we study the adoption of Mobile Agents for handling the VHE functionality. Mobile agents are nicely harmonized with the broader idea of VHE as they allow the autonomous execution of tasks by components that roam from node to node and network to network. We present the detailed modeling of a VHE provisioning architecture and investigate its suitability for different use cases and technical options (e.g., end user devices). The adoption of mobile agents for the ubiquitous provision of telecommunication services is quite promising in terms of efficiency. Through a series of experiments we quantify the performance benefits stemming from the adoption of mobile agents in contrast to conventional service provisioning schemes.     

An adaptive partitioning approach for mining discriminant regions in 3D image data

Mining discriminative spatial patterns in image data is an emerging subject of interest in medical imaging, meteorology, engineering, biology, and other fields. In this paper, we propose a novel approach for detecting spatial regions that are highly discriminative among different classes of three dimensional (3D) image data. The main idea of our approach is to treat the initial 3D image as a hyper-rectangle and search for discriminative regions by adaptively partitioning the space into progressively smaller hyper-rectangles (sub-regions). We use statistical information about each hyper-rectangle to guide the selectivity of the partitioning. A hyper-rectangle is partitioned only if its attribute cannot adequately discriminate among the distinct labeled classes, and it is sufficiently large for further splitting. To evaluate the discriminative power of the attributes corresponding to the detected regions, we performed classification experiments on artificial and real datasets. Our results show that the proposed method outperforms major competitors, achieving 30% and 15% better classification accuracy on synthetic and real data respectively while reducing by two orders of magnitude the number of statistical tests required by voxel-based approaches.