Human behaviour recognition in data-scarce domains

This paper presents the novel theory for performing multi-agent activity recognition without requiring large training corpora. The reduced need for data means that robust probabilistic recognition can be performed within domains where annotated datasets are traditionally unavailable. Complex human activities are composed from sequences of underlying primitive activities. We do not assume that the exact temporal ordering of primitives is necessary, so can represent complex activity using an unordered bag. Our three-tier architecture comprises low-level video tracking, event analysis and high-level inference. High-level inference is performed using a new, cascading extension of the Rao–Blackwellised Particle Filter. Simulated annealing is used to identify pairs of agents involved in multi-agent activity. We validate our framework using the benchmarked PETS 2006 video surveillance dataset and our own sequences, and achieve a mean recognition F-Score of 0.82. Our approach achieves a mean improvement of 17% over a Hidden Markov Model baseline.

Effective end wall profiling rules for a highly loaded compressor cascade

This paper presents a numerical study of a linear compressor cascade to investigate the effective end wall profiling rules for highly-loaded axial compressors. The first step in the research applies a correlation analysis for the different flow field parameters by a data mining over 600 profiling samples to quantify how variations of loss, secondary flow and passage vortex interact with each other under the influence of a profiled end wall. The result identifies the dominant role of corner separation for control of total pressure loss, providing a principle that only in the flow field with serious corner separation does the does the profiled end wall change total pressure loss, secondary flow and passage vortex in the same direction. Then in the second step, a multi-objective optimization of a profiled end wall is performed to reduce loss at design point and near stall point. The development of effective end wall profiling rules is based on the manner of secondary flow control rather than the geometry features of the end wall. Using the optimum end wall cases from the Pareto front, a quantitative tool for analyzing secondary flow control is employed. The driving force induced by a profiled end wall on different regions of end wall flow are subjected to a detailed analysis and identified for their positive/negative influences in relieving corner separation, from which the effective profiling rules are further confirmed. It is found that the profiling rules on a cascade show distinct differences at design point and near stall point, thus loss control of different operating points is generally independent.

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.

Evaluating the objective structured long examination record for nurse education

An objective structured long examination record (OSLER) is a modification of the long-case clinical examination and is mainly used in medical education. This study aims to obtain nursing students' views of the OSLER compared with the objective structured clinical examination (OSCE), which is used to assess discrete clinical skills. A sample of third-year undergraduate nursing students (n=21) volunteered to participate from a cohort of 230 students. Participants undertook the OSLER under examination conditions. Pre-and post-test questionnaires gathered the students' views on the assessments and these were analysed from a mainly qualitative perspective. Teachers' and simulated patient views were also used for data triangulation. The findings indicate that the OSLER ensures more holistic assessment of a student's clinical skills and particularly essential skills such as communication, and that the OSLER, together with the OSCE, should be used to supplement the assessment of clinical competence in nursing education.

Person Re-Identification using Deep Convnets with Multi-task Learning

Person re-identification involves recognizing a person across non-overlapping camera views, with different pose, illumination, and camera characteristics. We propose to tackle this problem by training a deep convolutional network to represent a person’s appearance as a low-dimensional feature vector that is invariant to common appearance variations encountered in the re-identification problem. Specifically, a Siamese-network architecture is used to train a feature extraction network using pairs of similar and dissimilar images. We show that use of a novel multi-task learning objective is crucial for regularizing the network parameters in order to prevent over-fitting due to the small size the training dataset. We complement the verification task, which is at the heart of re-identification, by training the network to jointly perform verification, identification, and to recognise attributes related to the clothing and pose of the person in each image. Additionally, we show that our proposed approach performs well even in the challenging cross-dataset scenario, which may better reflect real-world expected performance. 

The Status of Multi-Dimensional Core-Collapse Supernova Models

Models of neutrino-driven core-collapse supernova explosions have matured considerably in recent years. Explosions of low-mass progenitors can routinely be simulated in 1D, 2D, and 3D. Nucleosynthesis calculations indicate that these supernovae could be contributors of some lighter neutron-rich elements beyond iron. The explosion mechanism of more massive stars remains under investigation, although first 3D models of neutrino-driven explosions employing multi-group neutrino transport have become available. Together with earlier 2D models and more simplified 3D simulations, these have elucidated the interplay between neutrino heating and hydrodynamic instabilities in the post-shock region that is essential for shock revival. However, some physical ingredients may still need to be added/improved before simulations can robustly explain supernova explosions over a wide range of progenitors. Solutions recently suggested in the literature include uncertainties in the neutrino rates, rotation, and seed perturbations from convective shell burning. We review the implications of 3D simulations of shell burning in supernova progenitors for the ‘perturbations-aided neutrino-driven mechanism,’ whose efficacy is illustrated by the first successful multi-group neutrino hydrodynamics simulation of an 18 solar mass progenitor with 3D initial conditions. We conclude with speculations about the impact of 3D effects on the structure of massive stars through convective boundary mixing.