University of Macau

About: University of Macau is a education organization based out in Macao, Macau, China. It is known for research contribution in the topics: Computer science & Population. The organization has 6636 authors who have published 18324 publications receiving 327384 citations. The organization is also known as: UM & UMAC.

Multi-Head Attention with Disagreement Regularization

Abstract: Multi-head attention is appealing for the ability to jointly attend to information from different representation subspaces at different positions. In this work, we introduce a disagreement regularization to explicitly encourage the diversity among multiple attention heads. Specifically, we propose three types of disagreement regularization, which respectively encourage the subspace, the attended positions, and the output representation associated with each attention head to be different from other heads. Experimental results on widely-used WMT14 English-German and WMT17 Chinese-English translation tasks demonstrate the effectiveness and universality of the proposed approach.

Saponins from Chinese Medicines as Anticancer Agents

Abstract: Saponins are glycosides with triterpenoid or spirostane aglycones that demonstrate various pharmacological effects against mammalian diseases. To promote the research and development of anticancer agents from saponins, this review focuses on the anticancer properties of several typical naturally derived triterpenoid saponins (ginsenosides and saikosaponins) and steroid saponins (dioscin, polyphyllin, and timosaponin) isolated from Chinese medicines. These saponins exhibit in vitro and in vivo anticancer effects, such as anti-proliferation, anti-metastasis, anti-angiogenesis, anti-multidrug resistance, and autophagy regulation actions. In addition, related signaling pathways and target proteins involved in the anticancer effects of saponins are also summarized in this work.

Osteopontin -- a promising biomarker for cancer therapy.

Abstract: Osteopontin (OPN), a multifunctional protein, has emerged as a potentially valuable biomarker for diagnosing and treating cancers. Recent research focuses on its involvement in tumor biology including the cell proliferation, survival, angiogenesis, invasion, and metastasis. Understanding the molecular mechanisms and pharmacological effects of OPN in cancer development could lead to new targets for improving cancer diagnosis and treatment. This review explains how the structurally conserved domains of OPN are associated with OPN signaling mediators and CD44, and how the conserved OPN domains determine biological functions. The authors have reviewed representative works of OPN expression in breast cancer and colorectal cancer to elucidate the relationship between OPN and cancer/tumor biology. It has also been shown that the prognostic sensitivity in non-small cell lung cancer, hepatocellular carcinoma, gastric cancer, and ovarian cancer improved compared to the individual marker when OPN was analyzed in conjunction with other markers. The therapeutic approaches based on OPN inhibitors are discussed to illustrate recent research progress. Previous clinical data has indicated that OPN has played a unique role in cancer development, but further investigation is required to understand the underlying mechanism. More clinical trials are also required to examine the applicability and efficacy of OPN inhibitors in cancer therapy.

A Multi-Model Combination Approach for Probabilistic Wind Power Forecasting

Abstract: Short-term probabilistic wind power forecasting can provide critical quantified uncertainty information of wind generation for power system operation and control. It would be difficult to develop a universal forecasting model dominating over other alternative models because of the inherent stochastic nature of wind power. Therefore, a novel multi-model combination (MMC) approach for probabilistic wind power forecasting is proposed in this paper to exploit the advantages of different forecasting models. The proposed approach can combine different forecasting models those provide different kinds of probability density functions to improve the performance of probabilistic forecasting. Three probabilistic forecasting models based on the sparse Bayesian learning, kernel density estimation, and beta distribution fitting are used to form the combined model. The parameters of the MMC model are solved by two-step optimization. Comprehensive numerical studies illustrate the effectiveness of the proposed MMC approach.