1. 程式人生 > >林軒田《機器學習基石》課程總結

林軒田《機器學習基石》課程總結

最近釋出了一系列臺灣大學資訊工程系林軒田(Hsuan-Tien Lin)教授開設的《機器學習基石》的課程總結,分為4個部分,點選標題可檢視: 1. [機器什麼時候能夠學習?](https://www.cnblogs.com/analysis101/p/13479777.html)(When Can Machines Learn?) 2. [機器為什麼能夠學習?](https://www.cnblogs.com/analysis101/p/13511090.html)(Why Can Machines Learn?) 3. [機器怎樣學習?](https://www.cnblogs.com/analysis101/p/13672314.html)(How Can Machines Learn?) 4. [機器怎樣可以學得更好?](https://www.cnblogs.com/analysis101/p/13708180.html)(How Can Machines Learn Better?) ## 1 “3”的魔力 整個課程中,“3”這個數字貫穿始終。比如在介紹機器學習時,介紹了它和**3個相關領域**(資料探勘、人工智慧、統計學)的區別與聯絡: ![](https://upload-images.jianshu.io/upload_images/5715116-a693770b821c81c9.png?imageMogr2/auto-orient/strip%7CimageView2/2/w/1240) 在說明可學習性理論時,講了機器學習的**3個理論邊界**: ![](https://upload-images.jianshu.io/upload_images/5715116-3b0f02d1ed7acca9.png?imageMogr2/auto-orient/strip%7CimageView2/2/w/1240) 課程介紹了**3種線性模型**(PLA/Pocket、線性迴歸、邏輯迴歸): ![](https://upload-images.jianshu.io/upload_images/5715116-4af9d60679a9a8c7.png?imageMogr2/auto-orient/strip%7CimageView2/2/w/1240) 還介紹了**3種關鍵工具**(特徵變換、正則化、驗證): ![](https://upload-images.jianshu.io/upload_images/5715116-6e384880ae2ac614.png?imageMogr2/auto-orient/strip%7CimageView2/2/w/1240) 以及**3個學習原則**(奧卡姆剃刀、抽樣偏差、資料窺探): ![](https://upload-images.jianshu.io/upload_images/5715116-43de2e26d03838be.png?imageMogr2/auto-orient/strip%7CimageView2/2/w/1240) 在未來的《機器學習技法》課程中,還會**從3個方面**(更多特徵變換、更多正則化、更少標籤)進行**擴充套件**,進而引出一系列的模型: ![](https://upload-images.jianshu.io/upload_images/5715116-06675549d51e0dfe.png?imageMogr2/auto-orient/strip%7CimageView2/2/w/1240) ## 2 課程回顧 以下是每一節課的回顧。 ![](https://upload-images.jianshu.io/upload_images/5715116-063d62e74140dfcc.png?imageMogr2/auto-orient/strip%7CimageView2/2/w/1240) ![](https://upload-images.jianshu.io/upload_images/5715116-e6174ddec751977d.png?imageMogr2/auto-orient/strip%7CimageView2/2/w/1240) ![](https://upload-images.jianshu.io/upload_images/5715116-9524bced82934bbd.png?imageMogr2/auto-orient/strip%7CimageView2/2/w/1240) ![](https://upload-images.jianshu.io/upload_images/5715116-ec832bdcad5a4ab3.png?imageMogr2/auto-orient/strip%7CimageView2/2/w/1240) ![](https://upload-images.jianshu.io/upload_images/5715116-c217f4a6674024dc.png?imageMogr2/auto-orient/strip%7CimageView2/2/w/1240) ![](https://upload-images.jianshu.io/upload_images/5715116-e5c2d410b5bc830a.png?imageMogr2/auto-orient/strip%7CimageView2/2/w/1240) ![](https://upload-images.jianshu.io/upload_images/5715116-b3ab12d19e13d982.png?imageMogr2/auto-orient/strip%7CimageView2/2/w/1240) ![](https://upload-images.jianshu.io/upload_images/5715116-5375663fed499ac5.png?imageMogr2/auto-orient/strip%7CimageView2/2/w/1240) ![](https://upload-images.jianshu.io/upload_images/5715116-b08ba9b66926ad36.png?imageMogr2/auto-orient/strip%7CimageView2/2/w/1240) ![](https://upload-images.jianshu.io/upload_images/5715116-f31d8e8bc2f731cf.png?imageMogr2/auto-orient/strip%7CimageView2/2/w/1240) ![](https://upload-images.jianshu.io/upload_images/5715116-35b96e5d0fd25690.png?imageMogr2/auto-orient/strip%7CimageView2/2/w/1240) ![](https://upload-images.jianshu.io/upload_images/5715116-401d9ba060e6169f.png?imageMogr2/auto-orient/strip%7CimageView2/2/w/1240) ![](https://upload-images.jianshu.io/upload_images/5715116-5236b6ab033418e3.png?imageMogr2/auto-orient/strip%7CimageView2/2/w/1240) ![](https://upload-images.jianshu.io/upload_images/5715116-3f95d3ff6d871a6d.png?imageMogr2/auto-orient/strip%7CimageView2/2/w/1240) ![](https://upload-images.jianshu.io/upload_images/5715116-a2d6e39354db7766.png?imageMogr2/auto-orient/strip%7CimageView2/2/w/1240) ![](https://upload-images.jianshu.io/upload_images/5715116-8e9566b431ea227d.png?imageMogr2/auto-orient/strip%7CimageView2/2/