Shanghai Jiao Tong University

About: Shanghai Jiao Tong University is a education organization based out in Shanghai, Shanghai, China. It is known for research contribution in the topics: Population & Cancer. The organization has 157524 authors who have published 184620 publications receiving 3451038 citations. The organization is also known as: Shanghai Communications University & Shanghai Jiaotong University.

ReaxFF-lg: correction of the ReaxFF reactive force field for London dispersion, with applications to the equations of state for energetic materials.

Abstract: The practical levels of density functional theory (DFT) for solids (LDA, PBE, PW91, B3LYP) are well-known not to account adequately for the London dispersion (van der Waals attraction) so important in molecular solids, leading to equilibrium volumes for molecular crystals ∼10-15% too high. The ReaxFF reactive force field is based on fitting such DFT calculations and suffers from the same problem. In the paper we extend ReaxFF by adding a London dispersion term with a form such that it has low gradients (lg) at valence distances leaving the already optimized valence interactions intact but behaves as 1/R^6 for large distances. We derive here these lg corrections to ReaxFF based on the experimental crystal structure data for graphite, polyethylene (PE), carbon dioxide, and nitrogen and for energetic materials: hexahydro-1,3,5-trinitro- 1,3,5-s-triazine (RDX), pentaerythritol tetranitrate (PETN), 1,3,5-triamino-2,4,6-trinitrobenzene (TATB), and nitromethane (NM). After this dispersion correction the average error of predicted equilibrium volumes decreases from 18.5 to 4.2% for the above systems. We find that the calculated crystal structures and equation of state with ReaxFF-lg are in good agreement with experimental results. In particular, we examined the phase transition between α-RDX and γ-RDX, finding that ReaxFF-lg leads to excellent agreement for both the pressure and volume of this transition occurring at ∼4.8 GPa and ∼2.18 g/cm^3 density from ReaxFF-lg vs 3.9 GPa and ∼2.21 g/cm^3 from experiment. We expect ReaxFF-lg to improve the descriptions of the phase diagrams for other energetic materials.

Range Loss for Deep Face Recognition with Long-Tailed Training Data

Abstract: Deep convolutional neural networks have achieved significant improvements on face recognition task due to their ability to learn highly discriminative features from tremendous amounts of face images. Many large scale face datasets exhibit long-tail distribution where a small number of entities (persons) have large number of face images while a large number of persons only have very few face samples (long tail). Most of the existing works alleviate this problem by simply cutting the tailed data and only keep identities with enough number of examples. Unlike these work, this paper investigated how long-tailed data impact the training of face CNNs and develop a novel loss function, called range loss, to effectively utilize the tailed data in training process. More specifically, range loss is designed to reduce overall intrapersonal variations while enlarge interpersonal differences simultaneously. Extensive experiments on two face recognition benchmarks, Labeled Faces in the Wild (LFW) [11] and YouTube Faces (YTF) [33], demonstrate the effectiveness of the proposed range loss in overcoming the long tail effect, and show the good generalization ability of the proposed methods.