機器學習技法 (Machine Learning Techniques)

开始时间: 04/22/2022 持续时间: 8 weeks

所在平台: CourseraArchive

课程类别: 计算机科学

大学或机构: National Taiwan University(国立台湾大学)

授课老师: Hsuan-Tien Lin

课程主页: https://www.coursera.org/course/ntumltwo

课程评论: 2 个评论

评论课程        关注课程

课程详情

Welcome! The instructor has decided to teach the course in Mandarin on Coursera, while the slides of the course will be in English to ease the technical illustrations. We hope that this choice can help introduce Machine Learning to more students in the Mandarin-speaking world. The English-written slides will not require advanced English ability to understand, though. If you can understand the following descriptions of this course, you can probably follow the slides. [歡迎大家!這門課將採用英文投影片配合華文的教學講解,我們希望能藉這次華文教學的機會,將機器學習介紹給更多華人世界的同學們。課程中使用的英文投影片不會使用到艱深的英文,如果你能了解以下兩段的課程簡介,你應該也可以了解課程所使用的英文投影片。]

In the prequel of this course, Machine Learning Foundations, we have illustrated the necessary fundamentals that give any student of machine learning a solid foundation to explore further techniques. While many new techniques are being designed every day, some techniques stood the test of time and became popular tools nowadays.

The course roughly corresponds to the second half-semester of the National Taiwan University course "Machine Learning." Based on five years of teaching this popular course successfully (including winning the most prestigious teaching award of National Taiwan University) and discussing with many other scholars actively, the instructor chooses to focus on three of those popular tools, namely embedding numerous features (kernel models, such as support vector machine), combining predictive features (aggregation models, such as adaptive boosting), and distilling hidden features (extraction models, such as deep learning).


课程大纲

Each of the following items correspond to approximately one hour of video lecture. [以下的每個小項目對應到約一小時的線上課程]

Embedding Numerous Features [嵌入大量的特徵]
-- Linear Support Vector Machine [線性支持向量機]
-- Dual Support Vector Machine [對偶支持向量機]
-- Kernel Support Vector Machine [核型支持向量機]
-- Soft-Margin Support Vector Machine [軟式支持向量機]
-- Kernel Logistic Regression [核型羅吉斯迴歸]
-- Support Vector Regression
[支持向量迴歸]


Combining Predictive Features [融合預測性的特徵]
-- Bootstrap Aggregation [自助聚合法]
-- Adaptive Boosting [漸次提昇法]
-- Decision Tree [決策樹]
-- Random Forest [隨機森林]
-- Gradient Boosted Decision Tree [梯度提昇決策樹]

Distilling Hidden Features [萃取隱藏的特徵]
-- Neural Network [類神經網路]
-- Deep Learning [深度學習]
-- Radial Basis Function Network
[逕向基函數網路]
-- Matrix Factorization [矩陣分解]

Summary [總結]

课程评论(2条)

1

宋鑫要学习 2015-03-11 17:46 1 票支持; 0 票反对

1. 这门课是林轩田老师另一门课《机器学习基石》的后续课程,需要先学基石,再学技法;
2. 同Andrew Ng的机器学习课程相比,林老师的两门课更有深度,更重理论,趣味性上不输Ng的课;Ng的课强在注重实践,另外,Ng的课上有一些书本上不会写的ML黑魔法;
3. 课程视频不甚难,有一点本科微积分和统计的基础基本都能看懂;
4. 课后习题很重公式推导和程序实验,理论与实践并重。数学基础不好的人推导公式时会感觉十分吃力;程序实验题需要用到一些常用的机器学习包比如cvxopt,libsvm等,但是课程里并没有相应的引导,所以之前没接触过这些包的人也会十分吃力。 过年回家做作业做到半夜两点钟这种事情你以为我会告诉你吗

2

唐家声i 2015-02-24 09:00 2 票支持; 0 票反对

这门课算是结束了,因为太忙,证书拿得略带波折。
对这门课很有感情,因为第一门mooc证书就是一年前的机器学习基石,现在拿到了技法这门课的证书(等待最后发出来),也算是小圆满。
技法这门课,林老师讲授了很多有用的算法。本来讲到太多的算法,略有无头绪的感觉,但是这门课高质量的地方在于,林老师高度的概括和总结能力,几幅图,就把算法之间的关系和结构理清楚,按着这个结构,自己在头脑中过一遍全部内容,也会更加清晰。
作业方面,一如既往地难度高、压力大。很少会有mooc课程让我感受到这么大的作业压力。但是老师精心设计的作业,绝对会加深我们的理解。作业上,虽然有几个小错误,但是瑕不掩瑜。另外,第二次作业开始,每一次答题从以前的只给分数,变成了question level explaination了,这个设计着实让很多人松了口气...
林老师的课程理论一直很强,基石、技法皆如此。如果不只想做一个机器学习的爱好者,而是真正走进去的话,这两门课是不二选择。
最后,感谢林老师,没有这两门课,我也不会有机会获得现在的一些机会。
p.s. 看后期的论坛,发现drop掉的人还是蛮多的,这和基石第一次开课的时候一样。

课程简介

The course extends the fundamental tools in "Machine Learning Foundations" to powerful and practical models by three directions, which includes embedding numerous features, combining predictive features, and distilling hidden features. [這門課將先前「機器學習基石」課程中所學的基礎工具往三個方向延伸為強大而實用的工具。這三個方向包括嵌入大量的特徵、融合預測性的特徵、與萃取潛藏的特徵。]

课程标签

机器学习 机器学习技法 台湾大学 NTU 林轩田

105人关注该课程

主题相关的课程