Deep neural networks (DNNs) have the capacity to fit extremely noisy labels nonetheless they tend to learn data with clean labels first and then memorize those with noisy labels. We examine this behavior in light of the Shannon entropy of the predictions and demonstrate the low entropy predictions determined by a given threshold are much more reliable as the supervision than the original noisy labels. It also shows the advantage in maintaining more training samples than previous methods. Then, we power this entropy criterion with the Collaborative Label Correction (CLC) framework to further avoid undesired local minimums of the single network. A range of experiments have been conducted on multiple benchmarks with both synthetic and real-world settings. Extensive results indicate that our CLC outperforms several state-of-the-art methods.

No abstract

(1) Background: While previous studies revealed how underground mining might adversely affect the cardiopulmonary functions of workers, this study further investigated the differences between under- and aboveground mining at both high and low altitudes, which has received little attention in the literature. (2) Methods: Seventy-one healthy male coal mine workers were recruited, who had worked at least 5 years at the mining sites located above the ground at high (>3900 m; n = 19) and low (<120 m; n = 16) altitudes as well as under the ground at high (n = 20) and low (n = 16) altitudes. Participants’ heart rates, pulmonary functions, total energy expenditure and metabolism were measured over a 5-consecutive-day session at health clinics. (3) Results: Combining the results for both above- and underground locations, workers at high-altitude mining sites had significantly higher peak heart rate (HR), minimum average HR and training impulse as well as energy expenditure due to all substances and due to fat than those at low-altitude sites. They also had significantly higher uric acid, total cholesterol, creatine kinase and N-osteocalcin in their blood samples than the workers at low-altitude mining sites. At underground worksites, the participants working at high-altitude had a significantly higher average respiratory rate than those at low-altitude regions. (4) Conclusion: In addition to underground mining, attention should be paid to high-altitude mining as working under a hypoxia condition at such altitude likely presents physiological challenges.

Background Physical activity at pre-older ages (55–64 years) can greatly affect one’s physical fitness, health, physical-activity behaviour, and quality of life at older ages. The objective of this study was to conduct a 24-week walking-exercise programme among sedentary pre-older females and investigate the influence of different walking cadences on cardiorespiratory fitness and associated biomarkers. Methods A total of 78 pre-older sedentary female participants were recruited and randomly assigned to normal (n = 36), paced (n = 15), music-synchronised (n = 15) walking, and no-exercise control (n = 12) groups, respectively. The normal, paced, and music-synchronised walking groups walked at a cadence of 120 steps/min, 125 steps/min, and 120–128 steps/min, respectively, under supervised conditions. Anthropometric characteristics, step length, nutrient intake, blood pressure and composition, and cardiorespiratory fitness were measured at baseline, the 12th week of the programme, the 24th week of the programme (completion), and after a 12-week retention period, which began immediately upon completion of the programme and did not feature any supervised exercises. Results All walking conditions improved high-density lipoprotein cholesterol (HDL-C), low-density lipoprotein cholesterol, step length, maximum oxygen consumption (VO2max), and oxidative capacity at anaerobic threshold (all P < 0.001); however, after the 12-week retention period only the training effects of HDL-C (P < 0.05) and VO2max (P < 0.05) remained robust. Additionally, music-synchronised walking was found to reduce the fat ratio (P = 0.031), while paced walking was found to reduce body mass (P = 0.049). Conclusions The significant pre–post changes in health-related outcomes across the 24-week walking intervention, including improved blood composition, longer step length, and better cardiorespiratory capacity, show that this intervention is promising for improving health and fitness. When, during the retention period, the participants resumed their usual lifestyles without supervised exercise, most physiological biomarkers deteriorated. Thus, for sedentary middle-aged females, persistent behavioural change is necessary to retain the health benefits of physical exercise.