Sparse computation for large-scale binary classification

Well-known data mining algorithms rely on inputs in the form of pairwise similarities between objects. For large datasets it is computationally impossible to perform all pairwise comparisons. We therefore propose a novel approach that uses approximate Principal Component Analysis to efficiently identify groups of similar objects. The effectiveness of the approach is demonstrated in the context of binary classification using the supervised normalized cut as a classifier. For large datasets from the UCI repository, the approach significantly improves run times with minimal loss in accuracy.

Fully automatic segmentation of AP pelvis X-rays via random forest regression with efficient feature selection and hierarchical sparse shape composition

In clinical practice, traditional X-ray radiography is widely used, and knowledge of landmarks and contours in anteroposterior (AP) pelvis X-rays is invaluable for computer aided diagnosis, hip surgery planning and image-guided interventions. This paper presents a fully automatic approach for landmark detection and shape segmentation of both pelvis and femur in conventional AP X-ray images. Our approach is based on the framework of landmark detection via Random Forest (RF) regression and shape regularization via hierarchical sparse shape composition. We propose a visual feature FL-HoG (Flexible- Level Histogram of Oriented Gradients) and a feature selection algorithm based on trace radio optimization to improve the robustness and the efficacy of RF-based landmark detection. The landmark detection result is then used in a hierarchical sparse shape composition framework for shape regularization. Finally, the extracted shape contour is fine-tuned by a post-processing step based on low level image features. The experimental results demonstrate that our feature selection algorithm reduces the feature dimension in a factor of 40 and improves both training and test efficiency. Further experiments conducted on 436 clinical AP pelvis X-rays show that our approach achieves an average point-to-curve error around 1.2 mm for femur and 1.9 mm for pelvis.

Irrigation development in Eritrea: potentials and constraints: Proceedings of the workshop of AEAS and SLM Programme Eritrea

Development of irrigation, which is of crucial importance in Eritrea, is perceived by many as the main technique for improving the precarious food security situation in this Sahelian country in the Horn of Africa. The present publication presents the outcome of a nationwide workshop held in 2003, which brought together administrators, scientists, and members of public development agencies and NGOs. These workshop participants presented experiences, lessons learnt, and ideas about how to move forward in relation to development of irrigation in Eritrea. Specifically, the publication deals with the following broad themes, lessons learnt, and experiences in Eritrea: · spate irrigation systems and measurement of performance, as well as experience with modernisation of spate irrigation systems in Eritrea · small-scale irrigation systems and their potentials and pitfalls, including development of low-cost micro irrigation · climate and irrigation, including rainfall forecasts · socio-economic aspects of irrigation, including gender questions, institutional requirements, and irrigation and livelihoods The publication contains an extensive summary in the Tigrinya language, in order to facilitate access to the key findings by local non-English-speaking stakeholders in irrigation development.

Global Optimization with Sparse and Local Gaussian Process Models

We present a novel surrogate model-based global optimization framework allowing a large number of function evaluations. The method, called SpLEGO, is based on a multi-scale expected improvement (EI) framework relying on both sparse and local Gaussian process (GP) models. First, a bi-objective approach relying on a global sparse GP model is used to determine potential next sampling regions. Local GP models are then constructed within each selected region. The method subsequently employs the standard expected improvement criterion to deal with the exploration-exploitation trade-off within selected local models, leading to a decision on where to perform the next function evaluation(s). The potential of our approach is demonstrated using the so-called Sparse Pseudo-input GP as a global model. The algorithm is tested on four benchmark problems, whose number of starting points ranges from 102 to 104. Our results show that SpLEGO is effective and capable of solving problems with large number of starting points, and it even provides significant advantages when compared with state-of-the-art EI algorithms.

Object-Based Customization of Learning Content: Potentials and Pitfalls

Specification consortia and standardization bodies concentrate on e-Learning objects to en-sure reusability of content. Learning objects may be collected in a library and used for deriv-ing course offerings that are customized to the needs of different learning communities. How-ever, customization of courses is possible only if the logical dependencies between the learn-ing objects are known. Metadata for describing object relationships have been proposed in several e-Learning specifications. This paper discusses the customization potential of e-Learning objects but also the pitfalls that exist if content is customized inappropriately.

Differential effect of elevated intralabyrinthine pressure on ocular vestibular evoked myogenic potentials elicited by air conducted sound and bone conducted vibration

Recently, ocular vestibular evoked myogenic potentials (oVEMP) have emerged as a tool for assessment of utricular function. They are short-latency myogenic potentials which can be elicited in response to vestibular stimulation, e.g. by air-conducted sound (ACS) or bone-conducted vibration (BCV) (reviewed in (Kantner and Gurkov, 2012)). Otolithic afferent neurons trigger reflexive electromyographic activity of the extraocular muscles which can be recorded beneath the eye contralateral to the stimulated ear by use of surface electrodes.

Electrogram analysis and pacing are complimentary for recognition of abnormal conduction and far-field potentials during substrate mapping of infarct-related ventricular tachycardia.

BACKGROUND Mapping to identify scar-related ventricular tachycardia re-entry circuits during sinus rhythm focuses on sites with abnormal electrograms or pace-mapping findings of QRS morphology and long stimulus to QRS intervals. We hypothesized that (1) these methods do not necessarily identify the same sites and (2) some electrograms are far-field potentials that can be recognized by pacing. METHODS AND RESULTS From 12 patients with coronary disease and recurrent ventricular tachycardia undergoing catheter ablation, we retrospectively analyzed electrograms and pacing at 546 separate low bipolar voltage (<1.5 mV) sites. Electrograms were characterized as showing evidence of slow conduction if late potentials (56%) or fractionated potentials (76%) were present. Neither was present at (13%) sites. Pacing from the ablation catheter captured 70% of all electrograms. Higher bipolar voltage and fractionation were independent predictors for pace capture. There was a linear correlation between the stimulus to QRS duration during pacing and the lateness of a capturing electrogram (P<0.001), but electrogram and pacing markers of slow conduction were discordant at 40% of sites. Sites with far-field potentials, defined as those that remained visible and not captured by pacing stimuli, were identified at 48% of all pacing sites, especially in areas of low bipolar voltage and late potentials. Initial radiofrequency energy application rendered 74% of targeted sites electrically unexcitable. CONCLUSIONS Far-field potentials are common in scar areas. Combining analysis of electrogram characteristics and assessment of pace capture may refine identification of substrate targets for radiofrequency ablation.

Disturbances of agency and ownership in schizophrenia: An auditory verbal event related potentials study

A ‘sense of self’ is essentially the ability to distinguish between self-generated and external stimuli. It consists of at least two very basic senses: a sense of agency and a sense of ownership. Disturbances seem to provide a basic deficit in many psychiatric diseases. The aim of our study was to manipulate those qualities separately in 28 patients with schizophrenia (14 auditory hallucinators and 14 non-hallucinators) and 28 healthy controls (HC) and to investigate the effects on the topographies and the power of the event-related potential (ERP). We performed a 76-channel EEG while the participants performed the task as in our previous paper. We computed ERPs and difference maps for the conditions and compared the amount of agency and ownership between the HC and the patients. Furthermore, we compared the global field power and the topographies of these effects. Our data showed effects of agency and ownership in the healthy controls and the hallucinator group and to a lesser degree in the non-hallucinator group. We found a reduction of the N100 during the presence of agency, and a bilateral temporal negativity related to the presence of ownership. For the agency effects, we found significant differences between HC and the patients. Contrary to the expectations, our findings were more pronounced in non-hallucinators, suggesting a more profoundly disturbed sense of agency compared to hallucinators. A contemporary increase of global field power in both patient groups indicates a compensatory recruitment of other mechanisms not normally associated with the processing of agency and ownership.