Expected to be of the highest survey power telescope in the northern hemisphere,the Wide Field Survey Telescope(WFST)will begin its routine observations of the northern sky since 2023.WFST will produce a lot of scient...Expected to be of the highest survey power telescope in the northern hemisphere,the Wide Field Survey Telescope(WFST)will begin its routine observations of the northern sky since 2023.WFST will produce a lot of scientific data to support the researches of time-domain astronomy,asteroids and the solar system,galaxy formation and cosmology and so on.We estimated that the 5σlimiting magnitudes of WFST with 30 s exposure are u=22.31mag,g=23.42 mag,r=22.95 mag,i=22.43 mag,z=21.50 mag,w=23.61 mag.The above values are calculated for the conditions of airmass=1.2,seeing=075,precipitable water vapor=2.5 mm and Moon-object separation=45°at the darkest New Moon night of the Lenghu site(V=22.30 mag,Moon phaseθ=0°).The limiting magnitudes in different Moon phase conditions are also calculated.The calculations are based on the empirical transmittance data of WFST optics,the vendor provided CCD quantum efficiency,the atmospherical model transmittance and spectrum of the site.In the absence of measurement data such as sky transmittance and spectrum,we use model data.展开更多
随着电子音乐制作的普及,乐器数字接口(Musical Instrument Digital Interface,MIDI)信号的有效转换对音乐创作的意义日益凸显。为了提高电子音乐旋律自动转换的准确率,研究采用了一系列深度神经网络模型,对大型乐器数字接口数据集进行...随着电子音乐制作的普及,乐器数字接口(Musical Instrument Digital Interface,MIDI)信号的有效转换对音乐创作的意义日益凸显。为了提高电子音乐旋律自动转换的准确率,研究采用了一系列深度神经网络模型,对大型乐器数字接口数据集进行学习,并利用Adam算法进行优化。结果显示,在训练阶段和测试阶段,模型的识别率分别达到了95.03%和96.12%。这一结果不仅在训练集上超出了基线模型16.88%的表现,在测试集上的提升更是达到了18.36%,展现了模型优化后的强大泛化能力。在关键的性能指标上,模型的精确率为94.03%,召回率为95.06%,以及F1分数高达97.89%,这些指标均显著高于基线模型的原始得分。综合可以看出,基于深度学习的MIDI信号对电子音乐旋律自动转换研究,有效地提高了音乐转换的精度。展开更多
基金supported by the Strategic Priority Research Program of Chinese Academy of Sciences(grant Nos.XDB 41000000 and XDB 41010105)the National Science Foundation of China(NSFC,grant Nos.12233008,12173037 and 11973038)+1 种基金the China Manned Space Project(No.CMS-CSST-2021-A07)the Cyrus Chun Ying Tang Foundations。
文摘Expected to be of the highest survey power telescope in the northern hemisphere,the Wide Field Survey Telescope(WFST)will begin its routine observations of the northern sky since 2023.WFST will produce a lot of scientific data to support the researches of time-domain astronomy,asteroids and the solar system,galaxy formation and cosmology and so on.We estimated that the 5σlimiting magnitudes of WFST with 30 s exposure are u=22.31mag,g=23.42 mag,r=22.95 mag,i=22.43 mag,z=21.50 mag,w=23.61 mag.The above values are calculated for the conditions of airmass=1.2,seeing=075,precipitable water vapor=2.5 mm and Moon-object separation=45°at the darkest New Moon night of the Lenghu site(V=22.30 mag,Moon phaseθ=0°).The limiting magnitudes in different Moon phase conditions are also calculated.The calculations are based on the empirical transmittance data of WFST optics,the vendor provided CCD quantum efficiency,the atmospherical model transmittance and spectrum of the site.In the absence of measurement data such as sky transmittance and spectrum,we use model data.
文摘随着电子音乐制作的普及,乐器数字接口(Musical Instrument Digital Interface,MIDI)信号的有效转换对音乐创作的意义日益凸显。为了提高电子音乐旋律自动转换的准确率,研究采用了一系列深度神经网络模型,对大型乐器数字接口数据集进行学习,并利用Adam算法进行优化。结果显示,在训练阶段和测试阶段,模型的识别率分别达到了95.03%和96.12%。这一结果不仅在训练集上超出了基线模型16.88%的表现,在测试集上的提升更是达到了18.36%,展现了模型优化后的强大泛化能力。在关键的性能指标上,模型的精确率为94.03%,召回率为95.06%,以及F1分数高达97.89%,这些指标均显著高于基线模型的原始得分。综合可以看出,基于深度学习的MIDI信号对电子音乐旋律自动转换研究,有效地提高了音乐转换的精度。
