No abstract

In this article, we contribute to the earnings management literature by addressing the issue of Related Party Transactions (RPTs) during a firm's Initial Public Offering (IPO) process. We regard RPT-based earnings management as a kind of agency problem in the context of Chinese IPOs, and argue that the conflicts of interests between the controlling shareholders and the minority shareholders are the root of RPT-based earnings management in Chinese IPOs. We provide empirical evidence to demonstrate that RPT-based earnings management in a portfolio of earnings management tools including accruals management, and how it affects the firm's post-IPO long-term performance in China. Using 257 Chinese A and B shares IPOs during 1999 and 2000, our empirical findings suggest that controlling shareholders structure operating RPTs in pre-IPO period and these RPTs are positively associated with firm's operating performance. The decline in operating RPTs after IPO contributes to firm's post-IPO long-term underperformance and negatively affects firms' stock return.

Like other problems in computer vision, offline handwritten Chinese character recognition (HCCR) has achieved impressive results using convolutional neural network (CNN)-based methods. However, larger and deeper networks are needed to deliver state-of-the-art results in this domain. Such networks intuitively appear to incur high computational cost, and require the storage of a large number of parameters, which renders them unfeasible for deployment in portable devices. To solve this problem, we propose a Global Supervised Low-rank Expansion (GSLRE) method and an Adaptive Drop-weight (ADW) technique to solve the problems of speed and storage capacity. We design a nine-layer CNN for HCCR consisting of 3,755 classes, and devise an algorithm that can reduce the networks computational cost by nine times and compress the network to 1/18 of the original size of the baseline model, with only a 0.21% drop in accuracy. In tests, the proposed algorithm surpassed the best single-network performance reported thus far in the literature while requiring only 2.3 MB for storage. Furthermore, when integrated with our effective forward implementation, the recognition of an offline character image took only 9.7 ms on a CPU. Compared with the state-of-the-art CNN model for HCCR, our approach is approximately 30 times faster, yet 10 times more cost efficient.

No abstract

No abstract