Feature Selection Boosted by Unselected Features

Surrogate-Assisted and Filter-Based Multiobjective Evolutionary Feature Selection for Deep Learning

Feature selection (FS) for deep learning prediction models is a difficult topic for researchers to tackle. Most of the approaches proposed in the literature consist of embedded methods through the use of hidden layers added to the neural network architecture that modify the weights of the units associated with each input attribute so that the worst attributes have less weight in the learning process. Other approaches used for deep learning are filter methods, which are independent of the learning algorithm, which can limit the precision of the prediction model. Wrapper methods are impractical with deep learning due to their high computational cost. In this article, we propose new attribute subset evaluation FS methods for deep learning of the wrapper, filter and wrapper-filter hybrid types, where multiobjective and many-objective evolutionary algorithms are used as search strategies. A novel surrogate-assisted approach is used to reduce the high computational cost of the wrapper-type objective function, while the filter-type objective functions are based on correlation and an adaptation of the reliefF algorithm. The proposed techniques have been applied in a time series forecasting problem of air quality in the Spanish south-east and an indoor temperature forecasting problem in a domotic house, with promising results compared to other FS techniques used in the literature.

Altered brain functional connectivity in vegetative state and minimally conscious state

Objectives

The pathological mechanism for a disorder of consciousness (DoC) is still not fully understood. Based on traditional behavioral scales, there is a high rate of misdiagnosis for subtypes of DoC. We aimed to explore whether topological characterization may explain the pathological mechanisms of DoC and be effective in diagnosing the subtypes of DoC.

Methods

Using resting-state functional magnetic resonance imaging data, the weighted brain functional networks for normal control subjects and patients with vegetative state (VS) and minimally conscious state (MCS) were constructed. Global and local network characteristics of each group were analyzed. A support vector machine was employed to identify MCS and VS patients.

Results

The average connection strength was reduced in DoC patients and roughly equivalent in MCS and VS groups. Global efficiency, local efficiency, and clustering coefficients were reduced, and characteristic path length was increased in DoC patients (p < 0.05). For patients of both groups, global network measures were not significantly different (p > 0.05). Nodal efficiency, nodal local efficiency, and nodal clustering coefficient were reduced in frontoparietal brain areas, limbic structures, and occipital and temporal brain areas (p < 0.05). The comparison of nodal centrality suggested that DoC causes reorganization of the network structure on a large scale, especially the thalamus. Lobal network measures emphasized that the differences between the two groups of patients mainly involved frontoparietal brain areas. The accuracy, sensitivity, and specificity of the classifier for identifying MCS and VS patients were 89.83, 78.95, and 95%, respectively.

Conclusion

There is an association between altered network structures and clinical symptoms of DoC. With the help of network metrics, it is feasible to differentiate MCS and VS patients.

Smart Sports Outward Bound Training Assistant System Based on WSNs

The outward-bound training has been a popular manner to exercise in daily life. However, there lacks an intelligent assistant system to monitor the real-time status of users to avoid accidents during training. In order to fill this gap, this paper established an intelligent system to monitor fatigue status during outward-bound training by using surface electromyography (sEMG) signals. The system consists of three parts: a wearable device, edge node, and cloud server. First, the wearable device collects sEMG signals. Second, the edge node processes the collected sEMG signals and sends the sEMG signal features to the cloud server. Finally, the cloud server returns the results to edge node according to a stored classification model that learnt from massive historical sEMG signals. The experimental results show the effectiveness of the proposed system.

Kinect Body Sensor Technology-Based Quantitative Assessment Method for Basketball Teaching

Emphasizing the process and neglecting the end is the core idea of the research and implementation of college physical education learning and assessment, while the performance is the main form of evaluation results. This paper takes the quantitative assessment of basketball teaching as an example and proposes a new Kinect body sensor technology-based quantitative assessment method for basketball teaching. Specifically, for basketball technology recognition and assessment tasks, the Kinect body sensor is first used to collect volunteer's 3D skeleton motion data, then feeding the collected skeleton sequence to the vision transformer network to model the long-distance dependency. And based on this, the skeleton motion recognition network and skeleton motion assessment network are developed. The experimental results show that the proposed networks can well recognize and quantitatively assess the standard and non-standard basketball skill motions.