有人問我這個問題,想說就寫在這吧。目前還學不到家,只有斷斷續續讀個一陣子,有錯還請大家指正。
讀 Data mining 的書
一開始我讀 data mining 之類的課本、講義,發現內容講得滿概念的。但若只是想拿 ML 來用,那看這類文件會比較容易上手。比方 Introduction to Data Mining 寫得挺不錯的,可謂深入淺出!官網附的投影片挺棒的,可以先「快速」 (相對於看書來說……) 看一遍,再挑有興趣的部份閱讀課本。若有耐性慢慢讀完課本,會較有系統性地認識相關內容。由於碩論做 clustering 的緣故,我有好好地讀書本 clustering 部份。不過 clustering 論文種類多到爆炸,看完這本的內容也只是苦痛的開始而已……。
讀國外大學的授課講義
若想進一步了解 model 的原理,那就需要讀 ML 課程用的課本和講義。這類文件滿多的,其中我強力推薦 Andrew Ng 教的 Stanford CS229 Machine Learning。大部份教材在講解 model 最關鍵的數學時,只會從天空掉下來一句「套用 gradient ascending 求最大值」、「運用 EM 求最佳值」之類的話,接著結果就神奇地飛出來。這份講義卻從最基礎的數學開始講,包含需要用到的微分、線性代數、機率等,讓讀者有能力自己從頭推導數學算式,算是 ML 文件裡最親切的教材。當然,讀者也要有興趣慢慢地啃才行。我大概啃了一半多一點內容,收獲良多。可惜久一沒用也只記得概念了。有機會想再從頭自己推導一次。我覺得 Andrew Ng 數學寫得相當漂亮,這是我第一次察覺寫數學式和寫程式一樣,也有美醜、易懂難懂之分。附帶一提,當初會看這份教材是看 Mr. Saturday 推薦才讀的,感謝 Mr. Saturday 推薦。
後來我陸續看了一些不同國外大學的 ML 課程講義,發覺大家的切入點差很多,各有不同考量。像 MIT OCR 6.867 Machine Learning 有教 HMM,Stanford CS229 沒教。A Fundamentalist Organization of Machine Learning 提到由 Learning theory 的方式切入 ML,而不要走馬看花講一堆 model (作者稱為 “zoo approach”,挺傳神的譬喻)。我只有從 Andrew Ng 的講義裡讀了一點 learning theory,懂些基本觀念 (bias 和 variance),感覺挺有用的。若對 learning to theory 有興趣,印象中該文作者有推薦 15-859(B) Machine Learning Theory,Spring 2009。我看了一晚講義,內容是從簡單的假設環境中推導出明確的性質,還挺有趣的。但看到後來就看不懂了,也不明白能做什麼,就沒繼續看了。
讀 Wikipedia
此外,我發覺 Wikipedia 是學東西的好地方。有代表性的方法早已有系統地記錄下來,又提供詳細的參考資料可以一路讀下去。比方說想學 Linear classifier ,比起查書,查 Wikipedia 更有效率。我常拿 Wikipedia 內容和其它文件交叉對照,弄明白看不懂的部份。讀這些東西需要的是耐心以及不傷眼的 Wikipedia CSS 樣版。
讀經典課本
後來又看到 Mr. Saturday 提到 Elements of Statistical Learning: data mining,inference,and prediction. 2nd Edition.,而且官網還有附完整電子書,就載來看看。不得不說這本書數學超硬,但內容超級紮實,相當值得一讀。我挑有興趣且讀得下去的部份加減看,學到 bagging、random forest 等東西,讓我擴展眼界,明白許多原理。除數學太硬之外,這本書另一問題是實驗資料太小了,令人懷疑實驗結果是否適用於真實世界的情況。只好先暫時相信大師,之後再從實驗中確認。
有興趣的話可以在 Amazon 查一下其它書,或以這本書為底看相關書籍。ML 有不少經典書籍,也有針對影像處理或文字處理的。
其它
O’Reilly 出的 Programming Collective Intelligence - O’Reilly Media 也是不錯的書,適合初學者閱讀。配合使用情境,作者簡單地解釋 model 的概念並用 Python 說明如何實作。最令人讚賞的是,作者還有比較各個模型的優缺點。雖然分析的內容不見得正確,對初學者來說挺受用的。
總結
讀這些理論時,我發覺有許多文件提供不同的講解方式,可以多比較一番,找自己看得懂的。不用刻意從某個地方學會某種 model。文中提的是我學習的歷程,每個人的需求不同,數學底也不同,我的經驗不見得適用於其他人。寫在這裡供大家做個參考,有錯還請告知,感謝。
很棒的分享!謝謝︿︿
回覆刪除我目前也是碩一入門階段,持續努力中!
Good info
回覆刪除謝謝你的分享:)
回覆刪除I just got to this amazing site not long ago. I was actually captured with the piece of resources you have got here. Big thumbs up for making such wonderful blog page!
回覆刪除machine learning course
artificial intelligence course in mumbai
If I had to give a prime example of great quality content, this article would be one. It's well-written material that keeps your interest well.
回覆刪除SAP training in Kolkata
SAP training Kolkata
Best SAP training in Kolkata
SAP course in Kolkata
SAP training institute Kolkata
Thank you for this informative blog
回覆刪除python training in bangalore | python online training
aws training in bangalore | aws online training
artificial intelligence training in bangalore | artificial intelligence online training
machine learning training in bangalore | machine learning online training
blockchain training in bangalore | blockchain online training
uipath training in bangalore | uipath online training
Planning a hybrid event raises a host of new considerations; here are seven tips to help you prepare. Thank you, check also virtual edge and presenter bio template
回覆刪除It was wonerful reading your conent. Thankyou very much. # BOOST Your GOOGLE RANKING.It’s Your Time To Be On #1st Page
回覆刪除Our Motive is not just to create links but to get them indexed as will
Increase Domain Authority (DA).We’re on a mission to increase DA PA of your domain
High Quality Backlink Building Service
1000 Backlink at cheapest
50 High Quality Backlinks for just 50 INR
2000 Backlink at cheapest
5000 Backlink at cheapest
Thanks for the Valuable information.Really useful information. Thank you so much for sharing. It will help everyone.
回覆刪除SASVBA is one of the leading training providers in the country. We provide AI training Course in Delhi accordance with current industry standards, allowing students to find their dream job in the world's leading companies. SASVBA specializes in teaching services, artificial lighting training that imparts practical knowledge through live project-based learning.
FOR MORE INFO:
Event adult direction step player first. List will word.news headlines
回覆刪除Budget than others. Gun effort wind detail interest.top 10 news today
回覆刪除Thanks for the superb post.
回覆刪除artificial intelligence Classes in Pune
informative blog, keep posting full stack classes in pune
回覆刪除