从零学习机器学习_机器学习：如何从零变英雄

从零学习机器学习

以“为什么？”开头并以“我准备好了！”结尾 (Start with “Why?” and end with “I’m ready!”)

If your understanding of A.I. and Machine Learning is a big question mark, then this is the blog post for you. Here, I gradually increase your Awesomenessicity™ by gluing inspirational videos together with friendly text.

如果您对AI和机器学习的理解是一个很大的问号，那么这是适合您的博客文章。在这里，我通过将励志视频和友好的文本粘合在一起，逐渐提高您的Awesomenessicity ™。

Sit down and relax. These videos take time, and if they don’t inspire you to continue to the next section, fair enough.

坐下来放松一下。这些视频需要一些时间，如果它们不能激发您继续下一节的话，那还算公平。

However, if you find yourself at the bottom of this article, you’ve earned your well-rounded knowledge and passion for this new world. Where you go from there is up to you.

但是，如果您发现自己位于本文的底部，那么您已经获得了对这个新世界的全面了解和热情。从那里到哪里，取决于您。

了解为什么机器学习现在如此热门 (Understanding Why Machine Learning is so HOT Right Now)

A.I. was always cool, from moving a paddle in Pong to lighting you up with combos in Street Fighter.

AI总是很酷，从在Pong中移动桨到在Street Fighter中用连击点亮您。

A.I. has always revolved around a programmer’s functional guess at how something should behave. Fun, but programmers aren’t always gifted in programming A.I. as we often see. Just Google “epic game fails” to see glitches in A.I., physics, and sometimes even experienced human players.

AI一直围绕着程序员对某种事物应该如何表现的功能猜测。有趣，但是程序员并不总是像我们经常看到的那样对AI编程有天赋。只是Google的“史诗游戏”未能看到AI，物理学甚至有时是经验丰富的人类玩家的小故障。

Regardless, A.I. has a new talent. You can teach a computer to play video games, understand language, and even how to identify people or things. This tip-of-the-iceberg new skill comes from an old concept that only recently got the processing power to exist outside of theory.

无论如何，人工智能都有新的才能。您可以教计算机玩电子游戏，理解语言，甚至如何识别人或物。这种冰山一角的新技能来自一个旧概念，直到最近才使处理能力不存在于理论之外。

I’m talking about Machine Learning.

我说的是机器学习 。

You don’t need to come up with advanced algorithms anymore. You just have to teach a computer to come up with its own advanced algorithm.

您不再需要提出高级算法。您只需要教一台计算机配备其自己的高级算法即可。

So how does something like that even work? An algorithm isn’t really written as much as it is sort of… bred. I’m not using breeding as an analogy. Watch this short video, which gives excellent commentary and animations to the high-level concept of creating the A.I.

那么类似的东西怎么工作呢？实际上，算法的编写不如它的繁育。我没有将繁殖作为类比。观看这段简短的视频，它为创建AI的高级概念提供了出色的评论和动画

Wow! Right? That’s a crazy process!

哇！对？那是一个疯狂的过程！

Now how is it that we can’t even understand the algorithm when it’s done? One great visual was when the A.I. was written to beat Mario games. As a human, we all understand how to play a side-scroller, but identifying the predictive strategy of the resulting A.I. is insane.

现在怎么办，甚至无法理解算法呢？一种伟大的视觉效果是编写AI击败Mario游戏时。作为一个人类，我们都知道如何玩侧滚，但是确定最终人工智能的预测策略是疯狂的。

Impressed? There’s something amazing about this idea, right? The only problem is we don’t know Machine Learning, and we don’t know how to hook it up to video games.

印象深刻？这个想法有些不可思议，对吧？唯一的问题是我们不了解机器学习，也不知道如何将其连接到视频游戏。

Fortunately for you, Elon Musk already provided a non-profit company to do the latter. Yes, in a dozen lines of code you can hook up any A.I. you want to countless games/tasks! Check it out in action!

对您来说幸运的是，埃隆·马斯克 ( Elon Musk)已经提供了一家非营利性公司来做后者。是的，您可以在十几行代码中连接想要处理无数游戏/任务的任何AI！快来看看吧！

为什么要使用机器学习？ (Why Should You Use Machine Learning?)

I have two good answers on why you should care. Firstly, Machine Learning (ML) is making computers do things that we’ve never made computers do before. If you want to do something new, not just new to you, but to the world, you can do it with ML.

关于您为什么要关心我有两个很好的答案。首先，机器学习(ML)使计算机执行我们从未做过的事情。如果您想做一些新的事情，不仅是您自己，而是整个世界，您都可以使用ML来做。

Secondly, if you don’t influence the world, the world will influence you.

其次，如果您不影响世界，世界将影响您。

Right now significant companies are investing in ML, and we’re already seeing it change the world. Thought-leaders are warning that we can’t let this new age of algorithms exist outside of the public eye. Imagine if a few corporate monoliths controlled the Internet. If we don’t take up arms, the science won’t be ours. I think Christian Heilmann said it best in his talk on ML.

目前，重要的公司正在对ML进行投资，而且我们已经看到它改变了世界。思想领袖警告说，我们不能让这种新时代的算法存在于公众视野之外。想象一下，如果有几家公司垄断者控制着Internet。如果我们不采取行动，科学就不会成为我们的科学。我认为Christian Heilmann在有关ML的演讲中说得最好。

“We can hope that others use this power only for good. I — for one, don’t consider this a good bet. I’d rather play and be part of this revolution. And so can you.”

“我们可以希望其他人只能永远使用这种力量。我-一个，不要认为这是一个好选择。我更愿意参加这场革命。你也可以。”

好，现在我很感兴趣... (OK, now I’m interested…)

The concept is useful and cool. We understand it at a high level, but what the heck is actually happening? How does this work?

这个概念很有用又很酷。我们对此有较高的了解，但是到底发生了什么呢？这是如何运作的？

If you want to jump straight in, I suggest you skip this section and move on to the next “How Do I Get Started” section. If you’re motivated to be a DOer in ML, you won’t need these videos.

如果您想直接进入，建议您跳过本节，转到下一个“如何入门”部分。如果您有动机成为ML中的DOer，则不需要这些视频。

If you’re still trying to grasp how this could even be a thing, the following video is perfect for walking you through the logic, using the classic ML problem of handwriting.

如果您仍在尝试掌握这是怎么回事，那么以下视频非常适合通过经典的ML手写问题向您介绍逻辑。

Pretty cool huh? That video shows that each layer gets simpler rather than more complicated. Like the function is chewing data into smaller pieces that end in an abstract concept. You can get your hands dirty in interacting with this process on this site (by Adam Harley).

太酷了吧？该视频显示，每一层变得更简单而不是更复杂。就像该功能一样，将数据分成更小的片段，最后以抽象的概念结束。在此站点上与该过程进行交互时，您会很脏(由Adam Harley撰写 )。

It’s cool watching data go through a trained model, but you can even watch your neural network get trained.

看着数据经过训练有素的模型真是太酷了，但是您甚至可以看着您的神经网络受到训练。

One of the classic real-world examples of Machine Learning in action is the iris data set from 1936. In a presentation I attended by JavaFXpert’s overview on Machine Learning, I learned how you can use his tool to visualize the adjustment and back propagation of weights to neurons on a neural network. You get to watch it train the neural model!

实际的经典机器学习实例之一就是1936年的虹膜数据集。在我参加了JavaFXpert关于机器学习的概述的演讲中，我学习了如何使用他的工具可视化调整和反向传播。神经网络上神经元的权重您会看到它训练了神经模型！

Even if you’re not a Java buff, the presentation Jim gives on all things Machine Learning is a pretty cool 1.5+ hour introduction into ML concepts, which includes more info on many of the examples above.

即使您不是Java爱好者，Jim所提供的关于机器学习的所有内容的介绍也是 ML概念1.5个小时以上的超酷介绍，其中包括上述许多示例的更多信息。

These concepts are exciting! Are you ready to be the Einstein of this new era? Breakthroughs are happening every day, so get started now.

这些概念令人兴奋！您准备好成为这个新时代的爱因斯坦了吗？突破每天都在发生，所以现在就开始吧。

我该如何开始？ (How do I get started?)

There are tons of resources available. First, you should subscribe to some newsletters/twitter accounts to keep the personal hype train rolling. I started this one!

有大量可用资源。首先，您应该订阅一些新闻通讯/推特帐户，以保持个人炒作的节奏。我开始了这个！

Fun Machine Learning (@FunMachineLearn) | TwitterThe latest Tweets from Fun Machine Learning (@FunMachineLearn). Not for Machine Learning snobs. Enjoy the beauty and…twitter.com

有趣的机器学习(@FunMachineLearn)| Twitter 来自Fun Machine Learning(@FunMachineLearn)的最新推文。 不适用于机器学习势利小人。 享受美丽和… twitter.com

If you want some more high-level concepts, I suggest you take the non-technical course AI for Everyone on Coursera. This will get some terminology and examples in your brain as you adventure forward.

如果您需要更多高级概念，建议您在Coursera上参加针对所有人的非技术课程AI 。 当您前进时，这将在您的大脑中获得一些术语和示例。

As for “in-depth learning”, I’ll be recommending two approaches.

至于“深度学习”，我将推荐两种方法。

螺母n螺栓 (Nuts n Bolts)

In this approach, you’ll understand Machine Learning down to the algorithms and the math. I know this way sounds tough, but how cool would it be to really get into the details and code this stuff from scratch!

通过这种方法，您将了解机器学习以及算法和数学。我知道这种方式听起来很难，但是真正深入细节并从头开始编写这些东西会多么酷！

If you want to be a force in ML, and hold your own in deep conversations, then this is the route for you.

如果您想成为ML中的一员，并与自己进行深入的对话，那么这就是您的路。

I recommend that you try out Brilliant.org’s app (always great for any science lover) and take the Artificial Neural Network course. This course has no time limits and helps you learn ML while killing time in line on your phone.

我建议您尝试Brilliant.org的应用程序(对任何科学爱好者来说都非常好)，并参加“人工神经网络”课程。这门课程没有时间限制，可以帮助您学习ML，同时在手机上消磨时间。

This one costs money after Level 1.

1级后，这笔钱要花钱。

Combine the above with simultaneous enrollment in Andrew Ng’s Stanford course on “Machine Learning in 11 weeks”. This is the course that Jim Weaver recommended in his video above. I’ve also had this course independently suggested to me by Jen Looper.

将以上内容与同时参加安德鲁·伍 ( Andrew Ng )的斯坦福课程“ 11周机器学习”相结合。这是Jim Weaver在上面的视频中推荐的课程。詹·洛珀 ( Jen Looper)独立地建议我这门课。

Everyone provides a caveat that this course is tough. For some of you that’s a show stopper, but for others, that’s why you’re going to put yourself through it and collect a certificate saying you did.

每个人都警告说，这门课很难。对于你们中的某些人来说，这是一个表演的制止器，但是对于其他人来说，这就是为什么您要自己通过它并收集证明您做到了的证书的原因。

This course is 100% free. You only have to pay for a certificate if you want one.

本课程是100％免费的。您只需要支付一份证书就可以。

With those two courses, you’ll have a LOT of work to do. Everyone should be impressed if you make it through because that’s not simple.

有了这两门课程，您将有很多工作要做。如果一切顺利，每个人都应该印象深刻，因为那并不简单。

But more so, if you do make it through, you’ll have a deep understanding of the implementation of Machine Learning that will catapult you into successfully applying it in new and world-changing ways.

但是，更重要的是，如果您做到了这一点，您将对机器学习的实现有深刻的了解，这将使您成功地以崭新的，改变世界的方式成功地应用它。

极速赛车手 (Speed Racer)

If you’re not interested in writing the algorithms, but you want to use them to create the next breathtaking website/app, you should jump into TensorFlow and the crash course.

如果您对编写算法不感兴趣，但是想使用它们来创建下一个令人叹为观止的网站/应用程序，则应该跳入TensorFlow和速成课程。

TensorFlow is the de facto open-source software library for machine learning. It can be used in countless ways and even with JavaScript. Here’s a crash course.

TensorFlow是用于机器学习的事实上的开源软件库。它可以以无数种方式使用，甚至可以与JavaScript一起使用。这是速成班。

Plenty more information on available courses and rankings can be found here.

有关可用课程和排名的更多信息，请参见此处。

If taking a course is not your style, you’re still in luck. You don’t have to learn the nitty-gritty of ML in order to use it today. You can efficiently utilize ML as a service in many ways with tech giants who have trained models ready.

如果上课不是您的风格，那么您仍然很幸运。您不必今天就学习ML的精髓。您已经准备好训练模型的技术巨头可以通过多种方式有效地利用ML作为服务。

I would still caution you that there’s no guarantee that your data is safe or even yours, but the offerings of services for ML are quite attractive!

我仍然提醒您，不能保证您的数据甚至您的数据都是安全的，但是ML的服务吸引人！

Using an ML service might be the best solution for you if you’re excited and able to upload your data to Amazon/Microsoft/Google. I like to think of these services as a gateway drug to advanced ML. Either way, it’s good to get started now.

如果您很兴奋并且能够将数据上传到Amazon / Microsoft / Google，则使用ML服务可能是最适合您的解决方案。我喜欢将这些服务视为通往高级ML的门户药物。无论哪种方式，现在都可以开始。

更新！ (UPDATES!)

I created a 5 day intro to AI mini-course!!!

我创建了一个5天的AI迷你课程简介！！！

https://academy.infinite.red/p/ai-demystified-free-5-day-mini-course

Here are some awesome tutorials I’ve found which you should check out

这是我发现的一些很棒的教程，您应该查看

BrainJS tutorials — Neural Nets in JS
BrainJS教程-JS中的神经网络
TensorFlow tutorials code + video
TensorFlow教程代码+视频
Deep Learning Ocean — Kickstarter course
深度学习海洋-Kickstarter课程

让我们成为创造者 (Let’s Be Creators)

I have to say thank you to all the aforementioned people and videos. They were my inspiration to get started, and though I’m still a newb in the ML world, I’m happy to light the path for others as we embrace this awe-inspiring age we find ourselves in.

我必须对所有上述人员和视频表示感谢。他们是我起步的灵感，尽管我仍然是机器学习领域的新手，但我很高兴为我们拥抱自己所处的这个令人敬畏的时代，为其他人指明道路。

It’s imperative to reach out and connect with people if you take up learning this craft. Without friendly faces, answers, and sounding boards, anything can be hard. Just being able to ask and get a response is a game changer. Add me, and add the people mentioned above. Friendly people with friendly advice helps!

如果您要学习这项技术，就必须与人建立联系并与他人建立联系。没有友好的面Kong，答案和共鸣，任何事情都会变得很难。能够提出要求并得到回应的就是改变游戏规则的人。加我，并加上述人员。友善的人和友好的建议会有所帮助！

See?

看到？

I hope this article has inspired you and those around you to learn ML! I also would love for you to join me in finding cool and fun ML code. Star, watch, and contribute to my repo here: https://github.com/GantMan/fun-machine-learing

希望本文能启发您和您周围的人学习ML！我也很希望您能加入我的行列，找到有趣的ML代码。在这里加注星标，观看并为我的回购做贡献： https ： //github.com/GantMan/fun-machine-learing