Introduction to machine learning how to?

Machine learning and artificial intelligence is now the most popular learning direction,But everybody don't know how to start.今天,通过这篇文章,We have to tell everybody introduction to machine learning how to.

The first lesson of machine learning

The essence of machine learning is through the construction of the mathematical model framework,And depending on the machine itself constantly to optimize,Finally the optimal solution.因此,Our first lesson on the basis of the mathematical best,进行学习实践.

在整个学习过程中,You can have the following several kinds of mathematical knowledge：

1.线性代数：矩阵/张量乘法、求逆,奇异值分解/特征值分解,行列式,范数等

2.统计与概率：概率分布,独立性与贝叶斯,最大似然(MLE)和最大后验估计(MAP)等

3.优化：线性优化,非线性优化(凸优化/非凸优化)以及其衍生的求解方法如梯度下降、牛顿法、基因算法和模拟退火等

4.微积分：偏微分,链式法则,矩阵求导等

5.信息论、数值理论等

The mathematical theory for beginners,Is the threshold of a huge.因此,这里需要注意的点是：If mathematics basic to stop effect for you,Then put these things,From the ten algorithm of machine learning itself to learn,In the process of learning to,To make up for the inadequacy of their.After all, mathematics knowledge is only a tool of cognitive algorithm,而不是算法本身.

机器学习十大算法

作为最有名,And most important algorithm of machine learning ten,Although you don't say entirely cognitive,但是,You have to be to a deep understanding of what's inside.只有这样,To make you have a system of knowledge of machine learning and understanding.Here recommend zhou teacher《机器学习》一书,A classic a book.

At this stage in the learning process,You can meet the ten algorithm of：

(The Internet can find many data,Here I only listed the name,后续有机会,In each analysis of each algorithm for everyone)

1. 线性回归2. Logistic 回归3. 线性判别分析4. 分类和回归树5. 朴素贝叶斯6. K 最近邻算法7. 学习向量量化8. 支持向量机9. 袋装法和随机森林10. Boosting 和 AdaBoost

From the perspective of machine learning algorithm itself,Algorithm model is mainly divided into2个流派,A random forest“vote派”,简单来说,Is the sample file random block,Then respectively into the algorithm,With a high number of results as the final result.Another genre as“Feedback to send”,After get a result,Constant feedback to model,The parameters of the model through continuous,Final output optimal solution.

当然,在这个基础上,A lot of people are hard to calm down and learn one by one,Here you can recommend to watch algorithm,树结构,支持向量机,随机森林和Boosting.

Lie deep learning

Deep learning is the essence of constant feedback,And then deep learning model through constant adjustment parameters of feedback results,End up with an optimal solution.因此,With particular emphasis on a point is here,Deep learning itself is a kind of black box algorithm,Too much learning theory there is no any effect.你唯一能做的,Others have written on market is learning framework,Then try to adjust their parameters.And because the deep learning require very high performance of hardware,对于一般人来讲,Wasn't particularly friendly.

当然,Here is not to belittle deep learning algorithm,Here only stressed the point that,Deep learning algorithms are mostly exist in the method of adjustment parameters on,If the rapid adjustment parameters,达到最优的结果,What you need to do.

Machine learning related hardware and language selection

如果要做深度学习,Linux还是首选,因为其对很多学习模型支持比较好(主要是深度学习的Library).但即使你使用的是Windows系统,也可以用虚拟机装Ubuntu来进行学习.Deep learning model of small enough,Large deep learning we seldom locally/个人计算机上运行.至于编程语言,首推Python,Because of its good development supporting,主流的工具包都有Python版本.在特定情况下,选择R作为编程语言也是可以的.其他可能的语言还包括C++、Java和Matlab,但我个人不大推荐.Do not recommend,Mainly because of the language itself more dependent on the bottom of the environment,The performance is difficult to guarantee.

免费分享一些我整理的人工智能学习资料给大家,整理了很久,非常全面.包括一些人工智能基础入门视频+AI常用框架实战视频、图像识别、OpenCV、NLQ、YOLO、机器学习、pytorch、计算机视觉、深度学习与神经网络等视频、课件源码、国内外知名精华资源、AI热门论文等.

下面是部分截图,文末附免费下载方式.

目录

一、人工智能免费视频课程和项目

二、人工智能必读书籍

三、人工智能论文合集

四、机器学习+计算机视觉基础算法教程

 五、深度学习机器学习速查表（共26张）

学好人工智能,要多看书,多动手,多实践,要想提高自己的水平,一定要学会沉下心来慢慢的系统学习,最终才能有所收获.

点击下方名片,扫码免费下载文中资料.