机器学习最全知识点汇总
新机器视觉
共 1626字,需浏览 4分钟
·
2022-05-10 17:42
点击下方卡片,关注“新机器视觉”公众号
重磅干货,第一时间送达
1.列举常用的最优化方法
梯度下降法
牛顿法,
拟牛顿法
坐标下降法
梯度下降法的改进型如AdaDelta,AdaGrad,Adam,NAG等。
2.梯度下降法的关键点
![](https://filescdn.proginn.com/2a3f6df62e95c572f6a1eb448152fefb/26dae731960aaf737161284a71863a65.webp)
3.牛顿法的关键点
![](https://filescdn.proginn.com/85b3859df0b780f9e4a5c8371ccd7114/9152c1be099b45a2f63d93c25136e89c.webp)
![](https://filescdn.proginn.com/f3f2cca9c3b3a182dffb4a8e56bf18e1/ea70c9e8b92f8f856dffbe5233dba341.webp)
4.拉格朗日乘数法
![](https://filescdn.proginn.com/f5669115467d15922993d9d435b1b857/03e17fd0a9eda299b46a9854dca3c30d.webp)
![](https://filescdn.proginn.com/4d0b16424a17bc1d3f9bed5bf6cae346/3c072bcaed74a0fe200842a1ced6e37f.webp)
![](https://filescdn.proginn.com/7e5feaa2b7c7c6396c14d602f55b564e/1beb41919b8d79644a7ac1ea2dab0a3f.webp)
主成分分析
线性判别分析
流形学习中的拉普拉斯特征映射
隐马尔科夫模型
5.凸优化
优化变量的可行域是一个凸集
目标函数是一个凸函数
线性回归
岭回归
LASSO回归
Logistic回归
支持向量机
Softamx回归
6.拉格朗日对偶
![](https://filescdn.proginn.com/a3cd799fdd761e71b025309024258b9e/8f4f9455dde053f4da9949a00dbb629b.webp)
![](https://filescdn.proginn.com/646799266382a47e6408f9078dd4e066/19155279914f45a9036b03899d066232.webp)
![](https://filescdn.proginn.com/ff22860342716967aa44b9e0439499a5/49f079a3ca0685cf4b45236494092ff1.webp)
![](https://filescdn.proginn.com/71e61e549d0c27f7c73510c3c07119e8/ec19c0bffcfd427720134b150e2e0729.webp)
![](https://filescdn.proginn.com/d5fabcc7ef240000168419e6b6114123/95b14ee804e71570fa80cdaf562f4276.webp)
7.KKT条件
![](https://filescdn.proginn.com/700825b9ca9f81e3a8bd1c41096ebf6e/8f7e99bc42a3e86f3d7497b14c2ccca9.webp)
![](https://filescdn.proginn.com/cfce792e04d417a5fcb3810948c16991/ec91912d093efe38f0361c95805b7120.webp)
![](https://filescdn.proginn.com/cb71209da41f1fb25924e73af5a5038d/00f67f7817ac149766d122b9e83a0ee2.webp)
![](https://filescdn.proginn.com/f691026208c6fa8b0ff73332bfdafd5d/ef6e34628d8ee01ae75e4f96ebe49fc9.webp)
![](https://filescdn.proginn.com/6cfb35ebbe8cc0e6a28bb136b590aa38/14e36931d7d1cd6bc3c5a74677847358.webp)
![](https://filescdn.proginn.com/d98329138bf2eef695c9fae5ae157c30/339de3779b121c088d5ca1352c04c2da.webp)
![](https://filescdn.proginn.com/6302a9e800e5a0533b665dff64eb754e/002f2d754579d0d5ab753126a5e024ee.webp)
![](https://filescdn.proginn.com/2bfb3534b03c0b6776e1823fd5256a13/512e9d79cc4500d90d0db0a9d4ad4624.webp)
![](https://filescdn.proginn.com/684f7d6b5e3b5e4a90bb314d2bbd9f93/c5513a66f0bd7f7aeae98182a51a4b10.webp)
![](https://filescdn.proginn.com/12532cbb72980aec8a19c6179e36d9ae/e1b7e3d1de9c1ef230a3095665141405.webp)
8.特征值与特征向量
![](https://filescdn.proginn.com/cb71209da41f1fb25924e73af5a5038d/00f67f7817ac149766d122b9e83a0ee2.webp)
![](https://filescdn.proginn.com/b3964ec6857c4ff8fcbbb2e4a5ecd9d4/f44a62a4fba55ac7be7bbe5a40257d4e.webp)
![](https://filescdn.proginn.com/cb71209da41f1fb25924e73af5a5038d/00f67f7817ac149766d122b9e83a0ee2.webp)
![](https://filescdn.proginn.com/a7dff3e8f23407d25cc50a19701c3411/084f1480ffd80e08f04cdd90fc829248.webp)
![](https://filescdn.proginn.com/64b36d443abe5c6d071fed2aa249bd51/2717fcb65c32e9641d497c14a2960df2.webp)
![](https://filescdn.proginn.com/bf01c9a4266cfecc2e86596ca81e16d4/45aa80751918697b4dda400299b8be9e.webp)
![](https://filescdn.proginn.com/cde15110b3ce056e0c4e31aa7017ee33/c05e82b65b34796905f6915dc5d17bdb.webp)
9.奇异值分解
![](https://filescdn.proginn.com/378db19bcdd889a27eb0b6f3251c2828/f9ab7df7a3de90bc464a1dcfff108e74.webp)
![](https://filescdn.proginn.com/6de7b722661835019fe30d3273d16d73/53149b4c4cf488e71461b231ac6f8f2b.webp)
10.最大似然估计
![](https://filescdn.proginn.com/4a37922992407e629fe2004863c1b7ca/92b1af820e7ec2c40ade6fe5b598f2aa.webp)
![](https://filescdn.proginn.com/4a37922992407e629fe2004863c1b7ca/92b1af820e7ec2c40ade6fe5b598f2aa.webp)
![](https://filescdn.proginn.com/9fba7b413c5c1d7853cc1f773db61158/b9405804b09306cd4445c1bda24bcee2.webp)
![](https://filescdn.proginn.com/4a37922992407e629fe2004863c1b7ca/92b1af820e7ec2c40ade6fe5b598f2aa.webp)
![](https://filescdn.proginn.com/5371b3547716968ef7b6f51fd33c7f69/a35445b2ba6056462bcb23d4a37f123f.webp)
![](https://filescdn.proginn.com/4a37922992407e629fe2004863c1b7ca/92b1af820e7ec2c40ade6fe5b598f2aa.webp)
![](https://filescdn.proginn.com/2f8710a2f93c593820bafbbe0b105817/6eb72fa9195345d4a50d285c4ba635f0.webp)
![](https://filescdn.proginn.com/15e91f8ec7f66d69d02f83229fb89f31/b4b33179c3dfa950fd230d2aac0b5bf2.webp)
![](https://filescdn.proginn.com/9c1fa4d002bebc72f76478e670e52ff6/3a003bf12cad730c9c8090705ddd42b4.webp)
基本概念
1.有监督学习与无监督学习
2.分类问题与回归问题
3.生成模型与判别模型
![](https://filescdn.proginn.com/c27fdffb0bbb447e9ccf3489b1e4cee3/d91fd86c19598e6d40377a80ad6a5c71.webp)
4.交叉验证
5.过拟合与欠拟合
6.偏差与方差分解
![](https://filescdn.proginn.com/1ac226fbc11f9ca93ccb54b95e11c3c9/039b870d1ce28ea4591af23e809f715b.webp)
7.正则化
![](https://filescdn.proginn.com/4ce222033912dd2b4001ffebd005588e/891daec2777ca059e93652fbc17034f5.webp)
8.维数灾难
贝叶斯分类器
![](https://filescdn.proginn.com/18c3706eb566b50d96951f54a47f0c7b/736391aeb19e574f89b9679a124ce06f.webp)
![](https://filescdn.proginn.com/beb770a8622bac3e94cc3e8e881ac8ef/0e4d1c427a147c1d0d10171bfa67ab23.webp)
![](https://filescdn.proginn.com/5a1037b1c7b3eb28fab7cf5bf869a45a/5fdfe91454f13a3aa5b2f3adbe0eff55.webp)
决策树
1.用样本集D建立根节点,找到一个判定规则,将样本集分裂成D1和D2两部分,同时为根节点设置判定规则。
2.用样本集D1递归建立左子树。
3.用样本集D2递归建立右子树。
4.如果不能再进行分裂,则把节点标记为叶子节点,同时为它赋值。
![](https://filescdn.proginn.com/e125d507dadccb9b6b4063c3a711af33/6c2b1c016a97a82762ad85eebaa0a7ba.webp)
随机森林
AdaBoost算法
![](https://filescdn.proginn.com/dc3e6089953f77b663eca6f3fde1fcc9/0e2c5a3d814ebe74ddf363865712087e.webp)
![](https://filescdn.proginn.com/d7dc65f71f8a04d423cdc0e1e2c5595b/52908e25a2af399e08a565a9efd34c44.webp)
![](https://filescdn.proginn.com/a79480509b3c7a40b62cf71cb0c4b788/cf4e700de7742ea9998b5ab8b49f09de.webp)
![](https://filescdn.proginn.com/bd01e5f6dc23c2766b5e11f6e2cc10a1/e80f154b56645cce156a3cc421a9a6d9.webp)
主成分分析
![](https://filescdn.proginn.com/583983704638b12bd5ac7b99642b9a4c/5ba56a31addcb975da2b560f6761fb58.webp)
![](https://filescdn.proginn.com/ccab79ee80d9b840fc74e8c3d32a1800/87bc4ff66d6a54d1d3860daa86b39bf7.webp)
![](https://filescdn.proginn.com/e3ab61062a4dbcf49ba5cb8e427d0f60/207580cea6734402acb2a76adb1278f1.webp)
![](https://filescdn.proginn.com/a84cb97135d8a3b3e509d067a5781896/d95589ec9801a3cf51c1d890241db5c5.webp)
线性判别分析
![](https://filescdn.proginn.com/fd84606a3f9d4cc630ac534d89d2f685/faf36e50b46ef10a1d230c36fe06d3b2.webp)
![](https://filescdn.proginn.com/f7c6adb7eb486c70ba7aca923fef3dc3/33c179551f78badbfe945097ac127ed8.webp)
![](https://filescdn.proginn.com/28853cb44bb6d2384ce8fc600ee9daff/278a053a9954d29921e5131ee2454f4c.webp)
![](https://filescdn.proginn.com/9b804922fd0abc41e7d8fc7a40bc1ee4/e3056b8967298b6eaa7bc43d252b6eb9.webp)
![](https://filescdn.proginn.com/b568d399ec8d6cd88b79d0e091019b72/39ee52e47cad82e05ae200016b7a2709.webp)
kNN算法
人工神经网络
![](https://filescdn.proginn.com/08ca5aa6d805cbd8fb7ecb6a83f5c44c/4844edea81e598abe4408d241cc7c51f.webp)
![](https://filescdn.proginn.com/d51739f14ccce4c983eb88d16b388acf/984389ffbb2b0d0eabcd9d43f2e27b71.webp)
![](https://filescdn.proginn.com/e95c4a116d2a3c3e3223fe3ac2ec2625/470660b09305a00b6fbd2c0009120eef.webp)
![](https://filescdn.proginn.com/de69ee40ae6940f1220dfe6d1c8114b3/5d676094c7cab030c921feb8f8fbec72.webp)
支持向量机
![](https://filescdn.proginn.com/4ac49d2e0452fb0cf4c438f0bed2f296/d04b34871a190ca16c6e4bfad63c4b57.webp)
![](https://filescdn.proginn.com/eaae2b5b9ed6468c37163330e0834e35/afc901a88ded692e5e6b75ebc7b1d499.webp)
![](https://filescdn.proginn.com/ef5b0f152b64c77456a71bc8bfd7430c/4b0fb56c104c73e4daed407d3113096b.webp)
![](https://filescdn.proginn.com/704a2e1d32cfe0c877e97741e461ecb6/e77e523ee1283bd6d14c8be0d29e2851.webp)
logistic回归
![](https://filescdn.proginn.com/7b440b2cdc896ecfd739c180946f2997/e17813453e8ee19bad0e28b5a2ddf036.webp)
![](https://filescdn.proginn.com/15ab700665d049653089c37cc3783308/861532985d6077e1ef34a33e75bb4adc.webp)
![](https://filescdn.proginn.com/70416f272ca6b99e02dc69df587f0d6f/49182183f1cccc3bc18f578fed6ea6f4.webp)
![](https://filescdn.proginn.com/3848d0d3c5bce57894151440a617be7a/c6fadb34007bebece0e297575ca33406.webp)
![](https://filescdn.proginn.com/b4ba488d6e663deeddf128adc7e90ac7/cb7e8dcaf98f23314c884c7a11040288.webp)
K均值算法
卷积神经网络
![](https://filescdn.proginn.com/afbd0884c23a063ce8021a7072c20cb0/1cbf924ef012f204622a9bb3e11ba337.webp)
![](https://filescdn.proginn.com/5c60bca4da2fa8e81673b73635ff47c6/4092458916566768d7c68bdad000e9a5.webp)
![](https://filescdn.proginn.com/0006ac13cb38c49fecf35596b35ea601/fb131e94173493ab81fe705cc4d44bc8.webp)
![](https://filescdn.proginn.com/63ddb1c02d609b9a07ec1e4aa268a513/c975de6d51d7e8e77c9f50192d3fe728.webp)
![](https://filescdn.proginn.com/5aa0d61d852a1cf729ef7d6585fdef6b/20119b3d3f3a81030f8160a0b245235f.webp)
循环神经网络
![](https://filescdn.proginn.com/8832bcc32cecf472e4f3e404a104efeb/23f28a1ac887c7197d515c97c7bf0077.webp)
![](https://filescdn.proginn.com/5372c279858f46e34eaffd73a0557574/b0d1a7c34fb6f7d762bf01474d6dfe0f.webp)
![](https://filescdn.proginn.com/9856d947632407e5ed47d9b2e41b9d8c/5a378ef372cb3157fe7de68e19546231.webp)
![](https://filescdn.proginn.com/77e74add726771d7000882a496a4262b/f00633f12a8f6ff708e8b5fdbbb97b2b.webp)
![](https://filescdn.proginn.com/14f932ac02f01d448c76e3bd5509ae0c/f4806d62f8807bfab4ec094907e64215.webp)
高斯混合模型
![](https://filescdn.proginn.com/83a396f912ea56508ae0257ec8d0019a/48d2f9751bf627a4ac3a103e700ba1b6.webp)
![](https://filescdn.proginn.com/93747f5e5ecc94fb7997f9e6628dad93/7eafcba3122b234d40eee9f106da55c9.webp)
![](https://filescdn.proginn.com/34c3d2494052f388e74b036de42a7f23/622a940f9346fdb4218548e709428c5a.webp)
![](https://filescdn.proginn.com/0ee1e6cd3163ff5e0c8ce595e6b0dcc6/61d67904390ea14fffa66fdd089b62d8.webp)
![](https://filescdn.proginn.com/f1bf34a15677026d64e073cd220f98ee/d071b1bb760e49ca50071af6fccd9039.webp)
![](https://filescdn.proginn.com/e4cf2066bc490788a1bc05f5903923a3/b6f18816728d44b6d917ca071181a963.webp)
EM算法
![](https://filescdn.proginn.com/644f0f7b1a2399ac318355e9a96e6f54/27ad002acc41f62b08deeed9741bf429.webp)
![](https://filescdn.proginn.com/8a85baa27f7d0396890c87ab53202222/013fc34c7202fbd39b362ffc4c58588a.webp)
![](https://filescdn.proginn.com/6eb939b7e6e18526f6736032351a84a5/d72646031778919cb3748b47515d2284.webp)
![](https://filescdn.proginn.com/644f0f7b1a2399ac318355e9a96e6f54/27ad002acc41f62b08deeed9741bf429.webp)
![](https://filescdn.proginn.com/1ed4a4ae98ba0a95d737e652b8e900bf/80a3a4ede9ed9a55ecd87692a500582e.webp)
![](https://filescdn.proginn.com/54f1436d822f24290e2f5bbd3c44b859/e5adf4ef24c472a2227a600576a96f55.webp)
本文仅做学术分享,如有侵权,请联系删文。
评论