A Deep Neural Network Regularization Measure: The Class-Based Decorrelation Method.

In response to the challenge of overfitting, which may lead to a decline in network generalization performance, this paper proposes a new regularization technique, called the class-based decorrelation method (CDM). Specifically, this method views the neurons in a specific hidden layer as base learners, and aims to boost network generalization as well as model accuracy by minimizing the correlation among individual base learners while simultaneously maximizing their class-conditional correlation. Intuitively, CDM not only promotes diversity among the hidden neurons, but also enhances their cohesiveness among them when processing samples from the same class. Comparative experiments conducted on various datasets using deep models demonstrate that CDM effectively reduces overfitting and improves classification performance.

Development of a novel Ba2BiV3O11 photocatalyst with dual functions of both water oxidation and proton reduction performance.

To overcome the limitation of photocatalysts with dual functionality of water oxidation and proton reduction, we proposed a novel bismuth-based Ba2BiV3O11 (BBVO) photocatalyst, which can simultaneously drive the proton reduction reaction under UV light and water oxidation reaction under visible light. After doping with sulfur through an in situ vulcanization strategy, the light absorption and charge separation efficiencies of the sulfur-doped BBVO were significantly improved, thus boosting its oxygen evolution activity (64 µmol h-1) by more than 16 times compared with independent BBVO. The experimental results demonstrate that BBVO can be employed as a very promising bismuth-based photocatalyst for solar energy conversion.

Driving behavior characterization and traffic emission analysis considering the vehicle trajectory.

Based on the development of the concept of a resource-saving and environmentally friendly society, needing to develop low-carbon and sustainable urban transportation. Most of the pollutants come from the emissions of motor vehicle exhaust. Therefore, this paper analyzes the relationship between driving behavior and traffic emissions, to constrain driver behavior to reduce pollutant emissions. The GPS data are preprocessed by using Navicat for data integration, data screening, data sorting, etc., and then, the speed data are cleaned by using a combination of box-and-line plots and linear interpolation in SPSS. Second, this paper uses principal component analysis (PCA) to downsize 12 indicators such as average speed, average acceleration, and maximum speed and then adopts K-MEANS and K-MEDOIDS methods to cluster the driver's behavioral indicators, selects the aggregation method based on the clustering indexes optimally, and analyzes the driver's driving state by using the symbolic approximation aggregation method; finally, according to the above research results and combined with the MOVES traffic emission model to analyze the relationship between the driver's driving mode, driving state, and traffic emissions, the decision tree can be used to predict the unknown driving mode of the driver to estimate the degree of its emissions.