[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"$fM66f2Ojv6XF8jD9JquOgdc4-No8R342OljwUfU0cNdo":3},{"answer":4,"createTime":5,"id":6,"options":7,"origin":12,"question":18,"related":19,"source":29,"type":30},[],"2025-05-06 11:13:18",185832887,[8,9,10,11],"均方损失函数太复杂","均方损失非凸","均方损失没有偏导为0的点","均方损失没有极值点",{"count":13,"courseId":14,"courseImg":15,"courseName":16,"workId":17,"workName":16},85,"3ce5b2f686bbbaf3f2ef09732f063e06","https:\u002F\u002Ftihai-oss-cloud.itihey.com\u002Fimg\u002F51f99e5037c29fbba56117f952bfdad4.png","机器学习进步","exam_154527022","对数几率回归为什么不能通过令偏导为0求解",[20,31,40,49,52,61,70,79,88,97],{"answer":21,"createTime":5,"id":22,"options":23,"question":28,"source":29,"type":30},[],185832880,[24,25,26,27],"模型的泛化误差降低","模型的训练误差增加","模型的泛化误差增加","模型的计算速度增加","当假设空间中包含的模型复杂度增加时,通常会带来什么风险","v1",0,{"answer":32,"createTime":5,"id":33,"options":34,"question":39,"source":29,"type":30},[],185832883,[35,36,37,38],"根据不同阈值计算 TPR 和 FPR,并将结果绘制成曲线","直接使用分类模型的精确率和召回率绘制 ROC 曲线","ROC 曲线只与分类模型的最终预测类别相关,与概率预测无关","ROC 曲线中的面积等于模型的准确率","在构建 ROC 曲线时,模型的分类结果需要按某种方式计算真阳性率(TPR)和假阳性率 (FPR).以下哪项是正确的构建步骤",{"answer":41,"createTime":5,"id":42,"options":43,"question":48,"source":29,"type":30},[],185832885,[44,45,46,47],"频率主义","贝叶斯主义","人文主义","解构主义","统计学习属于什么主义",{"answer":50,"createTime":5,"id":6,"options":51,"question":18,"source":29,"type":30},[],[8,9,10,11],{"answer":53,"createTime":5,"id":54,"options":55,"question":60,"source":29,"type":30},[],185832889,[56,57,58,59],"RBF神经网络的激活函数是线性的","RBF神经网络中的隐藏层神经元采用的是高斯函数作为激活函数","RBF神经网络适用于处理非线性回归问题,但不适合分类任务","RBF神经网络的输出层采用的是Sigmoid激活函数.章节测试","下面关于RBF神经网络的说法正确的是",{"answer":62,"createTime":5,"id":63,"options":64,"question":69,"source":29,"type":30},[],185832892,[65,66,67,68],"两者更新权重的方法一样","两者更新权重的步幅一样","标准BP的效率低于累积BP","标准BP每一步更新的方向不同于累积BP,但是总体方向一致","关于标准BP和累积BP说法正确的是",{"answer":71,"createTime":5,"id":72,"options":73,"question":78,"source":29,"type":30},[],185832895,[74,75,76,77],"结构风险描述模型本身的某些性质","经验风险描述模型与训练数据的契合程度","经验风险对应于正则化项","正则化常数用于对结构风险和经验风险进行折中","下面有关正则化问题的说法错误的是()",{"answer":80,"createTime":5,"id":81,"options":82,"question":87,"source":29,"type":30},[],185832898,[83,84,85,86],"在训练集上表现优秀,但在测试集上表现较差","在训练集和测试集上表现均较差","模型对数据中的噪声不敏感","模型的复杂度较低","如果一个模型具有很高的方差,那么可能出现以下哪种情况",{"answer":89,"createTime":5,"id":90,"options":91,"question":96,"source":29,"type":30},[],185832900,[92,93,94,95],"间隔带两侧的松弛程度可有所不同","支持向量回归一般要求损失为 0 当且仅当模型的输出和实际值一样","支持向量回归也存在对偶问题","支持向量回归模型的解仍然具有稀疏性","下面关于支持向量回归, 说法错误的是",{"answer":98,"createTime":99,"id":100,"options":101,"question":106,"source":29,"type":30},[],"2025-05-06 11:13:19",185832902,[102,103,104,105],"表示定理表明,在核方法中,优化问题的解可以表示为样本点核函数的线性组合","表示定理适用于任意单调递增的正则化函数和非负损失函数的优化问题","表示定理仅适用于支持向量机和岭回归等少数几种机器学习模型","表示定理为核方法提供了理论基础,使得高维特征空间的操作可以通过核函数在低维空间实现","以下关于表示定理的描述中,哪一项是不正确的"]