Search

• We propose two kinds of multi-view Gaussian processes (GPs) frameworks called MvGP1 and MvGP2. • With the posterior consistency principle, our frameworks directly extend GPs to multi-view learning. • The MVGP1 significantly reduces the training time of the multi-view GPs with an acceptable accuracy. • The MVGP2 achieves state-of-the-art classification performance. Gaussian processes (GPs) are powerful tools for regression and classification tasks, and have been used widely in various fields of machine learning. However, GPs are barely applied directly to the scenario of multi-view learning so far. In this paper, we propose two kinds of multi-view GPs frameworks called MVGP1 and MVGP2 with posterior consistency, which combine multiple views by regularizing the marginal likelihood. Utilizing the posterior distribution consistency criterion, the MVGP1 is first presented. Then to further improve classification accuracy, we present the MVGP2, in which we introduce a trade-off parameter to govern the relative importance of the likelihood term compared with the regularization term. The MVGP1 significantly reduces the training time of the multi-view GPs with acceptable performance on the accuracy, and the MVGP2 achieves state-of-the-art classification performance. Experimental results on multiple real-world classification data sets show the effectiveness of the proposed frameworks. [ABSTRACT FROM AUTHOR]

Effects of heat stress (mean day/night temperatures of 35.2/16.1 and 27.4/15.6°C for heat stress and control, respectively) during different grain-filling phases (1–10, 11–20, 21–30, 31–40, and 1–40 days after pollination [DAP]) on the grain yield and quality of waxy maize were investigated using two varieties. Heat stress decreased the number and weight of grains, thereby reducing grain yield. The effects of heat stress at an early grain development stage (before 20 DAP) were severe. Compared with the control, starch deposition was not affected by heat stress in Suyunuo5, and it was only decreased by heat stress at 1–10 DAP in Yunuo7. Protein content responses to heat stress were variety and stage dependent. Heat stress increased the iodine-binding capacities of both varieties. Peak and breakdown viscosities were decreased by heat stress, and the effects gradually decreased with postponement of high temperature. Heat stress increased gelatinization temperature and retrogradation percentage for both varieties, and the response was stage dependent. Heat stress during the whole grain-filling period increased the pasting and gelatinization temperatures but decreased gelatinization enthalpy. Changes in protein and starch contents as well as the proportion of long chains in amylopectin affected the pasting and thermal properties.