-
Notifications
You must be signed in to change notification settings - Fork 0
/
树模型.drawio
16 lines (16 loc) · 3.72 KB
/
树模型.drawio
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
<mxfile host="app.diagrams.net" agent="Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/127.0.0.0 Safari/537.36 Edg/127.0.0.0" version="24.7.8">
<diagram name="第 1 页" id="6LExswBGzEWrd6V2PaQY">
<mxGraphModel dx="1356" dy="522" grid="1" gridSize="10" guides="1" tooltips="1" connect="1" arrows="1" fold="1" page="0" pageScale="1" pageWidth="827" pageHeight="1169" math="0" shadow="0">
<root>
<mxCell id="0" />
<mxCell id="1" parent="0" />
<mxCell id="NIyaHajtZH1n5Uw6bmwi-1" value="CART<div>(1)信息熵<span style="white-space: pre;">	</span><span style="white-space: pre;">	</span><span style="white-space: pre;">	</span><span style="white-space: pre;">	</span></div><div>(2)信息增益<span style="white-space: pre;">	</span><span style="white-space: pre;">	</span><span style="white-space: pre;">	</span><span style="background-color: initial;">ID3 <span style="white-space: pre;">	</span><span style="white-space: pre;">	</span>只能处理离散特征</span></div><div>(3)信息增益比<span style="white-space: pre;">	</span><span style="white-space: pre;">	</span><span style="white-space: pre;">	</span>C4.5<span style="white-space: pre;">	</span>能够处理连续特征</div><div>(4)基尼系数<span style="white-space: pre;">	</span><span style="white-space: pre;">	</span><span style="white-space: pre;">	</span>CART<span style="white-space: pre;">	</span></div><div>基尼系数不需要对数计算,高效一些,另外使用连续特征</div><div>(5)树的生成与修剪</div><div>引入正则项,重新遍历每个节点</div><div><br></div><div>(6)决策树优点:可解释性</div><div>(7)决策树缺点:可能生成复杂的树,过拟合(很容易在样本表现好)</div><div><br></div><div><br></div><div>(8)回归树 除了分裂标准,还要确定输出值</div><div>切分点选择(最小二乘法)输出值确定(样本均值)</div><div>步骤为:遍历特征与划分点,计算节点输出,根据样本计算损失函数</div><div><br></div><div><br></div><div>AdaBoost:每个弱分类器的权重不同,每轮迭代不同样本的权重不同</div><div>XGBoost:<span style="background-color: initial;">每个弱分类器的权重相同,每轮迭代不同样本的权重相同</span></div><div><br></div>" style="rounded=0;whiteSpace=wrap;html=1;align=left;verticalAlign=top;spacingLeft=4;" parent="1" vertex="1">
<mxGeometry x="260" y="210" width="420" height="350" as="geometry" />
</mxCell>
<mxCell id="MYtVJx9_-QZ0XqSOwh22-1" value="<div>DSSM最初也是nlp领域用来做语义相似度计算的,后面也成为了推荐领域很好用的模型。</div>" style="rounded=0;whiteSpace=wrap;html=1;align=left;verticalAlign=top;spacingLeft=4;" vertex="1" parent="1">
<mxGeometry x="-170" y="210" width="420" height="350" as="geometry" />
</mxCell>
</root>
</mxGraphModel>
</diagram>
</mxfile>