SemiHAR: Improving Semisupervised Human Activity Recognition via Multitask Learning.

Semisupervised human activity recognition (SemiHAR) has attracted attention in recent years from various domains, such as digital health and ambient intelligence. Currently, it still faces two challenges. For one thing, discriminative features may exist among multiple sequences rather than a single sequence since activities are combinations of motions involving several body parts. For another thing, labeled data and unlabeled data suffer from distribution discrepancies due to the different behavior patterns or biological conditions of users. For that, we propose a novel SemiHAR method based on multitask learning. First, a dimension-based Markov transition field (DMTF) technique is designed to generate 2-D activity data for capturing the interactions among different dimensions. Second, we jointly consider the user recognition (UR) task and the activity recognition (AR) task to reduce the underlying discrepancy. In addition, a task relation learner (TRL) is introduced to dynamically learn task relations, which enables the primary AR task to exploit preferred knowledge from other secondary tasks. We theoretically analyze the proposed SemiHAR and provide a novel generalization result. Extensive experiments conducted on four real-world datasets demonstrate that SemiHAR outperforms other state-of-the-art methods.

Distribution of heavy metals in coastal sediments under the influence of multiple factors: A case study from the south coast of an industrialized harbor city (Tangshan, China).

This research investigated the spatial distribution of heavy metals, including mercury (Hg), cadmium (Cd), copper (Cu), arsenic (As), nickel (Ni), lead (Pb), chromium (Cr), and zinc (Zn), in surface sediments from a coastal area near to an industrial harbor (Tangshan Harbor, China) using 161 sediment samples. According to the geo-accumulation index (Igeo), 11 samples were classified as unpolluted (Igeo≤0). Notably, 41.0 % of the research samples were moderately or strongly polluted (2 < Igeo≤3) with Hg and 60.2 % of the samples were moderately polluted (1 < Igeo≤2) in Cd. The ecological effect evaluation showed that the metals Zn, Cd, and Pb were at the effect range low level, and 51.6 % of the samples for Cu, 60.9 % for Cr, 90.7 % for As, 41.0 % for Hg, and 64.0 % for Ni fell in the range between the effect range-low and the effect range-mean levels, respectively. The correlation analysis showed that the distribution patterns of Cr, Cu, Zn, Ni, and Pb were similar to each other, high in the northwest, southeast, and southwest regions of the study area and low in the northeast region, which corresponded well with sediment size components. Based on principal component analysis (PCA) and positive matrix factorization (PMF), four distinct sources of pollution were quantitatively attributed, including agricultural activities (22.08 %), fossil fuel consumption (24.14 %), steel production (29.78 %), and natural sources (24.00 %). Hg (80.29 %), Cd (82.31 %) and As (65.33 %) in the region's coastal sediments were predominantly contributed by fossil fuel, steel production and agricultural sources, respectively. Cr (40.00 %), Cu (43.63 %), Ni (47.54 %), and Zn (38.98 %) were primarily of natural lithogenic origin, while Pb mainly came from the mixed sources of agricultural activities (36.63 %), fossil fuel (36.86 %), and steel production (34.35 %). Multiple factors played important roles in the selective transportation of sedimentary heavy metals, particularly sediment properties, and hydrodynamic sorting processes in the study area.

Shapelet Discovery by Lazy Time Series Classification.

As a representation of discriminative features, the time series shapelet has recently received considerable research interest. However, most shapelet-based classification models evaluate the differential ability of the shapelet on the whole training dataset, neglecting characteristic information contained in each instance to be classified and the classwise feature frequency information. Hence, the computational complexity of feature extraction is high, and the interpretability is inadequate. To this end, the efficiency of shapelet discovery is improved through a lazy strategy fusing global and local similarities. In the prediction process, the strategy learns a specific evaluation dataset for each instance, and then the captured characteristics are directly used to progressively reduce the uncertainty of the predicted class label. Moreover, a shapelet coverage score is defined to calculate the discriminability of each time stamp for different classes. The experimental results show that the proposed method is competitive with the benchmark methods and provides insight into the discriminative features of each time series and each type in the data.