Joint Analysis of Multiple Algorithms and Performance Measures

There has been an increasing interest in the development of new methods using Pareto optimality to deal with multi-objective criteria (for example, accuracy and time complexity). Once one has developed an approach to a problem of interest, the problem is then how to compare it with the state of art. In machine learning, algorithms are typically evaluated by comparing their performance on different data sets by means of statistical tests. Standard tests used for this purpose are able to consider jointly neither performance measures nor multiple competitors at once. The aim of this paper is to resolve these issues by developing statistical procedures that are able to account for multiple competing measures at the same time and to compare multiple algorithms altogether. In particular, we develop two tests: a frequentist procedure based on the generalized likelihood-ratio test and a Bayesian procedure based on a multinomial-Dirichlet conjugate model. We further extend them by discovering conditional independences among measures to reduce the number of parameters of such models, as usually the number of studied cases is very reduced in such comparisons. Data from a comparison among general purpose classifiers is used to show a practical application of our tests.

Uplink Performance of Conventional and Massive MIMO Cellular Systems with Delayed CSIT

This work studies the uplink of a cellular network with zero-forcing (ZF) receivers under imperfect channel state information at the base station. More specifically, apart from the pilot contamination, we investigate the effect of time variation of the channel due to the relative users' movement with regard to the base station. Our contributions include analytical expressions for the sum-rate with finite number of BS antennas, and also the asymptotic limits with infinite power and number of BS antennas, respectively. The numerical results provide interesting insights on how the user mobility degrades the system performance which extends previous results in the literature.

Bespoke Radiology e-Learning

Introduction: Foundation doctors are expected to assess and interpret plain x-ray studies of the chest/abdomen before a definitive report is issued by senior staff. The Royal College of Radiologists have published guidelines (RCR curriculum) on the scope of plain film findings medical students should be familiar with.1 Studies have shown that the x-ray interpretation without feedback does not significantly improve diagnostic ability. 2 Queen’s University, Belfast Trust Radiology and Experior Medical developed an online system to assess individual student ability to interpret X-ray findings. Over a series of assessments each student’s profile is built up, identifying strengths and weakness. The system can then create bespoke individual assessments re-evaluating previously identified weak areas and quantifying interpretative skill improvement. Aim: To determine how readily an online system is adopted by senior medical students, investigating if increasing exposure to x-ray interpretation combined with cyclical formative feedback enhances performance. Methods: The system was offered to all 270 final year medical students as an online resource. The system comprised a series of 20 weekly 30 minute assessments, containing normal and abnormal x-rays within the RCR curriculum. After each assessment students were given formative feedback, including their own result, annotated answers, peer group comparison and a breakdown of areas of strength and weakness. Focus groups of 4-5 students addressed student perspectives of the system, including ease of use, image resolution, system performance across different operating platforms, perceived value of formative feedback loops, breakdown of performance and the value of bespoke personalised assessments. Research Ethics Approval was granted for the study. Data analysis was via two-sided one-sample t-test; initial minimal recruitment was estimated as 60 students, to detect a mean 10% change in performance, with a standard deviation of 20%. Results and Discussion: Over 80% (n = XXX/270) of the student cohort engaged with the study. Student baseline average was 39%, increasing to 62% by the exit test. The steadily sustained improvement (57% relative performance in interpretative diagnostic accuracy) was despite increasing test difficulty. Student feedback via focus groups was universally positive throughout the examined domains. Conclusion: The online resource proved to be valuable, with high levels of student engagement, improving performance despite increasingly difficulty testing and positive learner experience with the system. References: 1. Undergraduate Radiology Curriculum, The Royal College of Ra, April 2012. Ref No. BFCR(12)4 The Royal College of Radiologists, April 2012 2. I Satia, S Bashagha, A Bibi, R Ahmed, S Mellor, F Zaman. Assessing the accuracy and certainty in interpretating chest x-rays in the medical division. Clin Med August 2013 Vol.13 no. 4 349-352