Minimizing the discrepancy of feature distributions between different domains is one of the most promising directions in unsupervised domain adaptation. From the perspective of moment matching, most existing discrepancy-based methods are designed to match the second-order or lower moments, which however, have limited expression of statistical characteristic for non-Gaussian distributions. In this work, we propose a Higher-order Moment Matching (HoMM) method, and further extend the HoMM into reproducing kernel Hilbert spaces (RKHS). In particular, our proposed HoMM can perform arbitrary-order moment matching, we show that the first-order HoMM is equivalent to Maximum Mean Discrepancy (MMD) and the second-order HoMM is equivalent to Correlation Alignment (CORAL). Moreover, HoMM (order≥ 3) is expected to perform fine-grained domain alignment as higher-order statistics can approximate more complex, non-Gaussian distributions. Besides, we also exploit the pseudo-labeled target samples to learn discriminative representations in the target domain, which further improves the transfer performance. Extensive experiments are conducted, showing that our proposed HoMM consistently outperforms the existing moment matching methods by a large margin. Codes are available at https://github.com/chenchao666/HoMM-Master

Molecular dynamics (MD) simulations are employed to study biaxial stretch-induced crystallization of polymers, during which the individual roles of chain conformation and orientation on crystal nucleation and growth are clarified. Systems with different stiffness and orientations are constructed by changing the stretch ratios of the x-and y-axis, which allow us to figure out the individual contributions of chain conformation and orientation to flowenhanced nucleation. The results show that nucleation occurs in areas with high segment orientation, and the higher orientation corresponds to the shorter nucleation induction period. The relationship between the nucleation induction period and orientation is quantitatively expressed, which indicates that orientation plays a dominant role in flow-enhanced nucleation. On the other hand, the results show that chain stiffness exhibits a negative correlation with nucleation in biaxial stretch, supporting that conformational entropy reduction is not the main driving force in flow-induced crystallization of polymers. With the secondary nucleation model, the crystal growth rates in different directions correlate well with the orientation at the growth front of the clusters, further confirming the decisive role of orientation in crystal nucleation and growth. Finally, crystal cluster merging is proposed to be a way to form shish structures in highly oriented melts.

No abstract

Ring polymers have the same critical adsorption point and crossover exponent as linear chains.