RESUMO
Underwater source localization by deep neural networks (DNNs) is challenging since training these DNNs generally requires a large amount of experimental data and is computationally expensive. In this paper, label distribution-guided transfer learning (LD-TL) for underwater source localization is proposed, where a one-dimensional convolutional neural network (1D-CNN) is pre-trained with the simulation data generated by an underwater acoustic propagation model and then fine-tuned with a very limited amount of experimental data. In particular, the experimental data for fine-tuning the pre-trained 1D-CNN are labeled with label distribution vectors instead of one-hot encoded vectors. Experimental results show that the performance of underwater source localization with a very limited amount of experimental data is significantly improved by the proposed LD-TL.
RESUMO
The significant deposition of tail fat in sheep has a profound impact on the economic benefits of animal husbandry. Furthermore, increasing the litter size is a crucial means of enhancing economic benefits. The BMPR1B and T/Brachyury genes are considered major functional genes that could affect sheep litter size and tail bone number, respectively. In this study, we employed direct sequencing to identify specific mutations of the BMPR1B gene in Gobi short tail sheep and carried out genotyping using MassARRAY technology for each variant of both the BMPR1B and T genes. Significant associations were demonstrated between the c.687G>A mutation of BMPR1B and the litter size in both the Gobi short tail sheep and Ujimqin sheep breeds. Meanwhile, the g.30058882_30058873GCAGATTAAAIndel mutation was significantly associated with the litter size in Gobi short tail sheep. These findings may provide valuable genetic markers for expanding sheep litter size. In addition, we also confirmed that the frequency of tail-bone-number-related T alleles was significantly higher in Gobi short tail sheep than in longer-tailed Ujimqin sheep.