[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"$f4wQy_3hw6ZdCIEdvmisAYap1nN-lYEEhdRVCc9Xa1vU":3},{"answer":4,"createTime":5,"id":6,"options":7,"origin":12,"question":15,"related":16,"source":26,"type":31},[],"2025-12-26 09:52:28",1076140886,[8,9,10,11],"对训练集随机采样,在随机采样的数据上建立模型","尝试使用在线机器学习算法","使用 PCA 算法减少特征维度","选项中没有正确答案",{"courseImg":13,"courseName":14},"https:\u002F\u002Ftihai-oss-cloud.itihey.com\u002Fimg\u002Fcd518c690d5b9dbacf33f77752e1df8d.jpg","机器学习","我们想要训练一个 ML 模型,样本数量有 100 万个,特征维度是 5000,面对如此大数据,如何有效地训练模型?( )",[17,28,32,41,50,59,67,75,80,89],{"answer":18,"createTime":5,"id":19,"options":20,"question":25,"source":26,"type":27},[],1076140877,[21,22,23,24],"K-Means","决策树","支持向量机","kNN","选项中哪些方法不可以直接来对文本分类?( )","v2",0,{"answer":29,"createTime":5,"id":6,"options":30,"question":15,"source":26,"type":31},[],[8,9,10,11],1,{"answer":33,"createTime":5,"id":34,"options":35,"question":40,"source":26,"type":27},[],1076140898,[36,37,38,39],"Ridge 回归适用于特征选择","Lasso 回归适用于特征选择","两个都适用于特征选择","选项中的说法都不对","关于特征选择,下列对 Ridge 回归和 Lasso 回归说法正确的是?( )",{"answer":42,"createTime":5,"id":43,"options":44,"question":49,"source":26,"type":27},[],1076140911,[45,46,47,48],"若 λ 较大时,偏差减小,方差减小","若 λ 较大时,偏差减小,方差增大","若 λ 较大时,偏差增大,方差减小","若 λ 较大时,偏差增大,方差增大","假如使用一个较复杂的回归模型来拟合样本数据,使用 Ridge 回归,调试正则化参数 λ,来降低模型复杂度.若 λ 较大时,关于偏差(bias)和方差(variance),下列说法正确的是?( )",{"answer":51,"createTime":5,"id":52,"options":53,"question":58,"source":26,"type":31},[],1076141013,[54,55,56,57],"纯度高的结点需要更多的信息来描述它","信息增益可以用"1比特-熵"获得","如果选择一个属性具有许多特征值, 那么这个信息增益是有偏差的","选项中说法都不对","在决策树分割结点的时候,下列关于信息增益说法正确的是( )",{"answer":60,"createTime":5,"id":61,"options":62,"question":66,"source":26,"type":31},[],1076141099,[63,64,65,57],"信息增益准则对可取值数目较多的属性有所偏好","增益率准则对可取值数目较少的属性有所偏好","C4.5算法并不是直接选择增益率最大的候选划分属性,而是先从候选划分属性中找出信息增益高于平均水平的属性,再从中选择增益率最高的","对于划分属性选择,选项中说法正确的是( )",{"answer":68,"createTime":5,"id":69,"options":70,"question":73,"source":26,"type":74},[],1076141119,[71,72],"对","错","在决策树学习过程中,如果当前结点划分属性为连续属性,那么该属性还可作为其后代结点的划分属性.( )",3,{"answer":76,"createTime":5,"id":77,"options":78,"question":79,"source":26,"type":74},[],1076141152,[71,72],"k均值算法可看作是高斯混合聚类在混合成分方差相等、且每个样本仅指派给一个混合成分时的特例.( )",{"answer":81,"createTime":5,"id":82,"options":83,"question":88,"source":26,"type":27},[],1076141221,[84,85,86,87],"C = 0","C = 1","C 正无穷大","C 负无穷大","假设我们使用原始的非线性可分版本的 Soft-SVM 优化目标函数.我们需要做什么来保证得到的模型是线性可分离的?( )",{"answer":90,"createTime":5,"id":91,"options":92,"question":96,"source":26,"type":27},[],1076141368,[93,94,95],"监督学习","非监督学习","半监督学习","智能化中医望诊时,对一幅舌脉图像(伸出舌头的人脸图像),希望把舌头部分从人脸的其他部分划分出来,可以采用以下方法:将整幅图的每个象素的属性记录在一张数据表中,然后用某种方法将这些数据按它们的自然分布状况划分成两类.因此每个象素就分别得到相应的类别号,从而实现了舌头图像的分割.那么这种方法属于:( )"]