DyCR: A Dynamic Clustering and Recovering Network for Few-Shot Class-Incremental Learning.

Few-shot class-incremental learning (FSCIL) aims to continually learn novel data with limited samples. One of the major challenges is the catastrophic forgetting problem of old knowledge while training the model on new data. To alleviate this problem, recent state-of-the-art methods adopt a well-trained static network with fixed parameters at incremental learning stages to maintain old knowledge. These methods suffer from the poor adaptation of the old model with new knowledge. In this work, a dynamic clustering and recovering network (DyCR) is proposed to tackle the adaptation problem and effectively mitigate the forgetting phenomena on FSCIL tasks. Unlike static FSCIL methods, the proposed DyCR network is dynamic and trainable during the incremental learning stages, which makes the network capable of learning new features and better adapting to novel data. To address the forgetting problem and improve the model performance, a novel orthogonal decomposition mechanism is developed to split the feature embeddings into context and category information. The context part is preserved and utilized to recover old class features in future incremental learning stages, which can mitigate the forgetting problem with a much smaller size of data than saving the raw exemplars. The category part is used to optimize the feature embedding space by moving different classes of samples far apart and squeezing the sample distances within the same classes during the training stage. Experiments show that the DyCR network outperforms existing methods on four benchmark datasets. The code is available at: https://github.com/zichengpan/DyCR.

Self-Supervised Lie Algebra Representation Learning via Optimal Canonical Metric.

Learning discriminative representation with limited training samples is emerging as an important yet challenging visual categorization task. While prior work has shown that incorporating self-supervised learning can improve performance, we found that the direct use of canonical metric in a Lie group is theoretically incorrect. In this article, we prove that a valid optimization measurement should be a canonical metric on Lie algebra. Based on the theoretical finding, this article introduces a novel self-supervised Lie algebra network (SLA-Net) representation learning framework. Via minimizing canonical metric distance between target and predicted Lie algebra representation within a computationally convenient vector space, SLA-Net avoids computing nontrivial geodesic (locally length-minimizing curve) metric on a manifold (curved space). By simultaneously optimizing a single set of parameters shared by self-supervised learning and supervised classification, the proposed SLA-Net gains improved generalization capability. Comprehensive evaluation results on eight public datasets show the effectiveness of SLA-Net for visual categorization with limited samples.

The executive pay gap and stock price crash risk: Promotion or suppression?

Background: In recent years, cases of stock price crash have continued to emerge. However, yet little research to date has investigated the compensation incentives of top management team (TMT) affect the risk of stock price crash. Nor has research considered the impact of the executive pay gap on the stock price crash risk. Especially, as the "egalitarianism" was broken in the compensation system, and the increase of the degree of marketization of salaries, the executive pay gap has shown an expanding trend. Under this circumstance, we would systematically examine the association between the extent of executive pay gap and its future stock price crash risk. Design methodology and approach: Based on the sample of A-Share non-financial listed companies in Shanghai and Shenzhen Stock Exchange, we used firm FE regression method to empirically examine the relationship of the internal and external compensation gaps of executives and crash risk, as well as its contigency variables and inner mechanism. Findings: The empirical results show that there is a U-shaped relationship between the internal and external pay gap of executives and future crash risk. After passing the endogenous test and the robustness test, the conclusion still holds. Further research shows that the U-shaped relationship between the pay gap and crash risk is more pronounced, when firms are affiliated with the non-state-owned enterprise or its compensation fairness is lower. Finally, the quality of information disclosure plays a mediation effect when executive pay gap affects stock price crash risk. Originality and value: According to the economic and behavior perspectives, we explored the impact of compensation structure on stock price crash risk from the pay gap of executives for the first time, and extended the emerging literature of forecasting future stock price crash risk and executive pay gap. In addition, a key implication of our findings is that more guidance for firms is provided to design the compensation structures and to reduce stock price crash risk.

Can the social trust promote corporate green innovation? Evidence from China.

Corporate green innovation has played a crucial role in balancing profitability and environmental protection. The existing research on determinant factors of green innovation has its main defects in emphasizing excessively enterprise's formal institutional environment and neglecting the informal institutional environment, causing an incomplete understanding of the relationship between institutional environments and corporate green innovation. To bridge this gap, using a sample of Shanghai and Shenzhen A-share listed firms in manufacturing industry during the period of 2010-2016, we investigate how social trust, an important informal institutions, affects corporate green innovation. Our results show that social trust is positively associated with green innovation, remaining valid after applying endogenous and robustness tests. In addition, the positive relationship between social trust and green innovation is more prominent when the enterprise is non-state-owned or locates in a looser command-and-control (CAC) environmental regulations region. Further analysis shows that social trust boosts corporate green innovation via promoting knowledge sharing, decreasing financing constraints, and fulfilling more corporate social responsibility (CSR). This paper enriches the literature concerning social trust and green innovation and draws back more public attention on the role of informal institutions play in promoting green innovation.

Harder than diamond: superior indentation strength of wurtzite BN and lonsdaleite.

Recent indentation experiments indicate that wurtzite BN (w-BN) exhibits surprisingly high hardness that rivals that of diamond. Here we unveil a novel two-stage shear deformation mechanism responsible for this unexpected result. We show by first-principles calculations that large normal compressive pressures under indenters can compel w-BN into a stronger structure through a volume-conserving bond-flipping structural phase transformation during indentation which produces significant enhancement in its strength, propelling it above diamond's. We further demonstrate that the same mechanism also works in lonsdaleite (hexagonal diamond) and produces superior indentation strength that is 58% higher than the corresponding value of diamond, setting a new record.

Colossal shear-strength enhancement of low-density cubic BC2N by nanoindentation.

Recently synthesized low-density cubic BC2N exhibits surprisingly high shear strength inferred by nanoindentation in stark contrast to its relatively low elastic moduli. We show by first-principles calculation that this intriguing phenomenon can be ascribed to a novel structural hardening mechanism due to the compressive stress beneath the indenter. It significantly strengthens the weak bonds connecting the shear planes, yielding a colossal enhancement in shear strength. The resulting biaxial stress state produces atomistic fracture modes qualitatively different from those under pure shear stress. These results provide the first consistent explanation for a variety of experiments on the low-density cubic BC2N phase across a large range of strain.