[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"$f-9stEwFa-oKgYEw81pWvgCfz80MTC1Gm0Y9VSCUGGqU":3},{"answer":4,"createTime":5,"id":6,"options":7,"origin":12,"question":18,"related":19,"source":30,"type":31},[],"2024-05-09 08:54:23",986807720,[8,9,10,11],"它是凸函数,凸函数无法解决非凸问题","它可以有负值","它无法配合交叉熵损失函数使用","当输入值过大或者过小时,梯度趋近于0,容易造成梯度消失问题",{"courseId":13,"courseImg":14,"courseName":15,"workId":16,"workName":17},"1000076607","https:\u002F\u002Ftihai-oss-cloud.itihey.com\u002Fimg\u002Fd7ea7086329261ddbccb0e9e00c955b1.jpg","[智慧共享课]人工智能引论","57250198","第六章单元测试","关于sigmoid激活函数,下列描述正确的是( )",[20,32,40,49,52],{"answer":21,"createTime":22,"id":23,"options":24,"question":29,"source":30,"type":31},[],"2024-05-09 08:54:22",986807394,[25,26,27,28],"BP算法是一种将输出层误差反向传播给隐藏层进行参数更新的方法","BP算法将误差从后向前传递,获得各层单元所产生误差,进而依据这个误差来让各层单元修正各单元参数","对前馈神经网络而言,BP算法可调整相邻层神经元之间的连接权重大小","在BP算法中,每个神经元单元可包含不可偏导的映射函数","下面对误差反向传播 (error back propagation, BP)描述不正确的是( )","v2",0,{"answer":33,"createTime":5,"id":34,"options":35,"question":38,"source":30,"type":39},[],986807640,[36,37],"对","错","激活函数的引入和增强模型的非线性拟合能力.( )",3,{"answer":41,"createTime":5,"id":42,"options":43,"question":48,"source":30,"type":31},[],986807671,[44,45,46,47],"是一种端到端学习的方法","是一种监督学习的方法","实现了非线性映射","隐藏层数目大小对学习性能影响不大","下面对前馈神经网络这种深度学习方法描述不正确的是( )",{"answer":50,"createTime":5,"id":6,"options":51,"question":18,"source":30,"type":31},[],[8,9,10,11],{"answer":53,"createTime":5,"id":54,"options":55,"question":60,"source":30,"type":61},[],986807767,[56,57,58,59],"在梯度下降和随机梯度下降中,为了最小化损失函数,通常使用循环迭代的方式不断更新模型参数","在每次迭代中,随机梯度下降需要计算训练集所有样本的误差和,用于更新模型参数","在每次迭代中,梯度下降使用所有数据或者部分训练数据,用于更新模型参数","梯度下降是遗传算法的一种参数优化算法","以下关于梯度下降和随机梯度下降的说明,哪些描述是正确的( )",1]