ABSTRACT
Training deep graph neural networks (GNNs) is notoriously hard. Besides the standard plights in training deep architectures such as vanishing gradients and overfitting, it also uniquely suffers from over-smoothing, information squashing, and so on, which limits their potential power for encoding the high-order neighbor structure in large-scale graphs. Although numerous efforts are proposed to address these limitations, such as various forms of skip connections, graph normalization, and random dropping, it is difficult to disentangle the advantages brought by a deep GNN architecture from those "tricks" necessary to train such an architecture. Moreover, the lack of a standardized benchmark with fair and consistent experimental settings poses an almost insurmountable obstacle to gauge the effectiveness of new mechanisms. In view of those, we present the first fair and reproducible benchmark dedicated to assessing the "tricks" of training deep GNNs. We categorize existing approaches, investigate their hyperparameter sensitivity, and unify the basic configuration. Comprehensive evaluations are then conducted on tens of representative graph datasets including the recent large-scale Open Graph Benchmark, with diverse deep GNN backbones. We demonstrate that an organic combo of initial connection, identity mapping, group and batch normalization attains the new state-of-the-art results for deep GNNs on large datasets. Codes are available: https://github.com/VITA-Group/Deep_GCN_Benchmarking.
ABSTRACT
CaTiO3 is considered to be one of the most promising catalysts for the degradation of organic pollutants, but its application is limited by the wide band gap and low catalytic activity. Element doping is an effective strategy to solve these problems. Herein, a novel CaTiO3 co-doped with Ag and Co (Ca1-xAgxTi1-yCoyO3) was synthesized by combining co-precipitation and the microwave hydrothermal method for the first time. The crystal structure, microstructure and light absorption of the material were systematically investigated. The results showed that Ca1-xAgxTi1-yCoyO3 had higher light absorption than pure CaTiO3, and the band gap was reduced to 2.78 eV. First-principles calculations indicated that Ag-Ca and Co-Ti tended to form donor-acceptor defect pairs in the doping process. These defect states not only enhanced the adsorption properties, but also could be used as carrier traps to optimize the dielectric properties of CaTiO3. In the photoelectrocatalytic system, with 0.01 g of catalyst, 98% of methylene blue in 100 mL solution (10 mg L-1) was degraded in 150 min. In addition, Ca1-xAgxTi1-yCoyO3 showed strong stability and excellent recyclability. The double ion co-doping technology will provide an effective strategy for improving the catalytic activity of traditional wide-band gap semiconductors.