机器学习是一种算法么
Recently, there has been a lot of fuss about the word Learning. We always hear about machine learning, deep learning, neural networks, etc. But what does this really mean? Did scientists find a way to create a brain-like component and implement it inside machines? Is it just a marketing word used to sell software and services? Are machines going to take over the world? What on earth is this about?
最近，关于学习一词有很多大惊小怪的事情。 我们总是听到有关机器学习 ， 深度学习 ， 神经网络等的信息。但这到底意味着什么？ 科学家是否找到了创建类似于大脑的组件并将其在机器内部实现的方法？ 它只是用于销售软件和服务的营销词汇吗？ 机器会占领世界吗？ 这到底是什么？
Well… do not worry, we are still not there… yet! In this article I will introduce what learning really is, and how scientists teach computers to achieve human-like tasks, or in some cases, even outperform humans. This article will try to address this issue in the most non-tech way I can imagine. It is written, so that anyone from any domain will be able to grasp the idea.
好吧...别担心，我们还不在那里... 呢！ 在本文中，我将介绍什么是真正的学习，以及科学家如何教计算机实现类似于人类的任务，甚至在某些情况下甚至胜过人类。 本文将尝试以我能想到的最非技术性的方式解决此问题。 它的写作是为了使来自任何领域的任何人都可以理解这个想法。
为什么我们需要学习？ (Why do we need Learning?)
Before learning took place, problem-solving tasks relied on writing algorithms. An algorithm is simply a set of rules that takes an input and returns an output as a solution for the problem.
在进行学习之前，解决问题的任务依赖于编写算法 。 算法只是一组规则，这些规则接受输入并返回输出作为问题的解决方案。
Consider the following: Given a list of numbers, you are asked to sort them in an increasing order. This problem is solved by algorithms. In fact, there are a lot of algorithms to solve this task. They work by taking the list, applying some rules and manipulations, and returning the list in a sorted way. This problem and other similar ones were “somehow” easy for computer scientists. They only had to think and come up with an algorithm to solve a given task.
请考虑以下情况：给定一个数字列表，要求您按升序对它们进行排序。 该问题通过算法解决。 实际上，有很多算法可以解决此任务。 他们通过获取列表，应用一些规则和操作并以排序方式返回列表来工作。 对于计算机科学家来说，这个问题和其他类似的问题“很容易”解决。 他们只需要思考并提出一种算法即可解决给定的任务。
On the other hand, some problems were not so easy to solve by algorithms. People started to ask more from computers. They wanted the machine to have super abilities of solving very hard tasks. Tasks that scientists completely had no idea how to program. For example: How is it possible to write an algorithm, that takes an image of an animal and outputs the type of it? This is a very easy task for humans, but solving it with algorithms is a complex mission if not impossible. Humans know how to classify animal photos, but they do not know how to describe the steps they take to reach the answer. Thus, an important question arises — How to solve problems that even humans do not know how to describe?
另一方面，有些问题不是那么容易用算法解决。 人们开始向计算机提出更多要求。 他们希望机器具有解决非常艰巨任务的超级能力。 科学家完全不知道如何编程的任务。 例如：如何编写一种算法，以获取动物的图像并输出其类型？ 对于人类而言，这是一项非常容易的任务，但是即使不是不可能的事情，使用算法来解决也是一项复杂的任务。 人类知道如何对动物照片进行分类，但是他们不知道如何描述他们为获得答案所采取的步骤。 因此，出现了一个重要的问题-如何解决甚至人类都不知道如何描述的问题？
算法世界！ (The world of Algorithms!)
On the internet, algorithms are all around you. You’re reading this article because an algorithm brought it to you, among others, to click which you did, and the algorithm took note when you opened it. The algorithm decides what you see when you search through your photos and does the finding, and maybe even makes a little movie for you.
在互联网上，算法无处不在。 您正在阅读本文，是因为一种算法将其带给您，其中包括单击所做的操作，并且在打开时该算法会进行记录。 该算法决定您搜索照片并进行发现时所看到的内容，甚至可以为您制作一部小电影。
When you buy something, the algorithm sets the price, and the algorithm is at your bank watching transactions for fraud. The stock market is full of algorithms, trading with algorithms. Given this, you might want to know how these little algorithmic body shaping your world work, especially when they did in the olden days human-built algorithmic bots by giving them instructions that humans could explain in the form of “if this, then that”, but many problems are just too big and hard for a human to write simple instructions for, like finding out a fraudulent transaction in a billion financial transactions a second.
当您购买东西时，该算法确定价格，并且该算法在您的银行监视交易是否欺诈。 股市充满了算法，需要进行算法交易。 鉴于此，您可能想知道这些小的算法主体如何塑造您的世界，特别是当它们在过去的人类人造算法机器人中所做的工作时，向他们提供人类可以以“如果这样，那么那样”的形式进行解释的说明，但是许多问题对于人类而言却难以编写简单的说明，例如，在一秒钟的十亿次金融交易中找出欺诈性交易。
Consider an Airline. For this airline seat, what is the maximum price? The price that the user will pay right now is determined by the algorithmic bot. Not perfect answers but much better than a human could do! No one knows, not even the humans who built them or will build them in future knows how they exactly do it!
考虑航空公司。 对于这个飞机座位，最高价是多少？ 用户现在要支付的价格由算法机器人确定。 不是完美的答案，但是比人类能做到的要好得多！ 没有人知道，甚至建造它们或将来将要建造它们的人都不知道他们是如何做到的！
人工算法失败 (Failure of Human Written Algorithms)
What is in the following picture? Is it a Bee, or is it a 3? For humans, even little humans, its easy to say, but it’s impossible to just tell a bot in bot language, which is the language of logic gates, as to how to do it.
下图是什么？ 是蜜蜂，还是3只？ 对于人类，即使是很小的人类，它也很容易说出来，但是要做到这一点，就不可能仅用一种机器人语言(一种逻辑门语言)来告诉机器人。
Really we just know that’s a bee and that’s a 3. We can say in words what makes them different but a bot doesn't understand words and it’s the wiring in our brains that makes it happen.
真的，我们只知道那是一只蜜蜂，而那只是3只。我们可以用言语来表达使它们与众不同的原因，但是机器人无法理解言语，而正是大脑中的联系使之得以实现。
While an individual neuron in our brain may be understood, clusters of neurons which form the brain are hard to grasp. The whole is beyond our understanding, none the less, it works. So to get a bot that can do this sorting between bees and threes, you don’t build it yourself, you build a bot that builds bots and a bot that teaches bots. These bots brains are simpler, something a smart human programmer can make.
虽然可以理解我们大脑中的单个神经元，但难以掌握形成大脑的神经元簇。 整体而言，它仍然是我们无法理解的。 因此，要获得一个可以在蜜蜂和三分之二之间进行这种分类的机器人，您就不必自己构建它，而是要构建一个可以构建机器人的机器人和一个可以教机器人的机器人。 这些机器人的大脑更简单，这是聪明的人类程序员可以做的事情。
Remember, the builder bot builds dots, but the teacher bot doesn't teach, only tests the student bots(why does this sound familiar!?).
记住，构建器机器人会构建点，但教师机器人不会教，只会测试学生机器人(为什么听起来很熟悉！？)。
Student bots are not very good at it, actually. At first, builder bots makes random connections in random bots brains. This leads to some very special student bots sent to teacher bot to be taught to distinguish the bee from the number three. The human gives teacher bot a bunch of bee photos and three photos, and an answer key to which is what.
实际上，学生机器人并不是很好。 首先，构建器机器人会在随机机器人的大脑中建立随机连接。 这导致一些非常特殊的学生机器人被发送到教师机器人，以教导他们将蜜蜂与三号机器人区分开。 人类给老师机器人一堆蜜蜂照片和三张照片，以及答案的关键。
The student bots take the test, but they are bad at what they do. This not their fault, they were built that way. They’re brought back to Builder bot, and those that did best are put to one side, say one who scored at least 50%, and the others, recycled. Builder bot still isn’t good at building bots, but now it takes those left and makes copies with changes in new combination, and sent back to school. They go, teacher bot tests again and Builder bot builds, the average test score raises again to say, 60%, again and again and again until 99.99% perfection. So what we’ve done till now is that, before, humans could write algorithms for their machines and bots, but now, machines create their own algorithms for tasks
学生机器人参加了测试，但是他们做的不好。 这不是他们的错，他们是那样建造的。 他们被带回Builder bot，而表现最好的机器人则被放到一边，例如得分至少为50％的人，而其他人则被回收。 Builder bot仍然不擅长构建bot，但现在它会将剩下的东西带走，并以新组合的更改进行复制，然后送回学校。 他们去了，老师机器人再次进行测试，而Builder bot进行了构建，平均测试分数一次又一次地提高了60％，直到达到99.99％的完美为止。 因此，到目前为止，我们要做的是，人类可以为自己的机器和机器人编写算法，但是现在，机器可以为任务创建自己的算法
So a trial and error approach is done, and no, this is not done for one or two bots, its done for a billion different bots with different random connections, and the test questions are another billion pages long. As mentioned, this is done for problems which we cannot tell the bots what is what in bot language.
因此，我们尝试并尝试了一种方法， 但不是，对于一个或两个机器人来说 ， 这是不可行的，对于十亿个具有不同随机连接的机器人来说，它是不可行的，测试问题又长了十亿页。 如前所述，这样做是为了解决一些我们无法告诉机器人以机器人语言表示的问题。
Teacher bot isn’t over seeing an old schoolhouse with a dozen students, but an infinite warehouse with thousands of students, and the test isn’t 10 questions, but a million question, and how many times do they test?
机器人机器人不是在看一个有十几个学生的老校舍，而是一个有成千上万学生的无限仓库，测试不是10个问题，而是100万个问题，他们测试了多少次？
Builder bot repeats the building process(randomized connections). At first students that survived are just lucky. But by combining enough lucky bots and keeping only what works and randomly messing around with new copies of that, eventually a student bot emerges that isn’t lucky, and that can perhaps barely tell bees from threes. This spot is copied and the change is applied over to all other bots too. Slowly the average test score rises, and the grade needed to survive the next round gets higher and higher — from 50%, to 60%, 70%, 80%, 90%, 95%, etc.
Builder bot重复构建过程(随机连接)。 起初，幸存下来的学生很幸运。 但是，通过组合足够多的幸运机器人并仅保留有效的方法，并随机将其与新的副本混为一谈，最终会出现一个不幸运的学生机器人，它可能几乎无法分辨蜜蜂。 复制该位置，并将更改也应用于所有其他机器人。 平均考试成绩慢慢上升，下一轮生存所需的等级越来越高-从50％升至60％，70％，80％，90％，95％等。
Keep this loop up and eventually, from the infinite schoolhouse, a student bot will emerge, which can tell a bee from the three in a photo, with an accuracy of 99.999%. That number means that in a hypothetical collection of 1,000,000 photographs, it gets only 1 photo wrong! It’s never seen before pretty well, but how the student bot does this, neither the teacher bot, the builder bot or even the human overseer can understand, nor the student by itself. After keeping so many useful random changes the wiring and its head is incredibly complicated, and while an individual line of code may be understood and clusters of codes can’t be easily understood. The whole is beyond our understanding, nonetheless.
保持循环，最终从无限的校舍中出现一个学生机器人，该机器人可以在照片中分辨出这三者中的一只蜜蜂，其准确性为99.999％。 这个数字意味着在假设的1,000,000张照片中，只有1张照片是错误的！ 以前从未见过如此好，但是学生机器人如何做到这一点，老师机器人，建筑机器人甚至人类监督者都无法理解，学生本身也无法理解。 进行了许多有用的随机更改后，接线及其头部变得异常复杂，并且尽管可能会理解一行代码，但不容易理解代码簇。 尽管如此，整体还是超出了我们的理解范围。
It works, but this is frustrating, especially as the student bot is very good at exactly only the kinds of questions it has been taught. It’s great with photos, but useless with videos or baffled if the photos are upside down or things that are obviously not bees, like a cat. Our teacher bot can’t teach everything, all the human overseer can do is give it more questions to make the test even longer to include the kinds of questions even the best bots get wrong. This is important to understand. It’s a reason why companies are obsessed with collecting data — more data equals longer tests, which equals better bots. When you get the are you human test on a website, you are not only proving that you are human, but hopefully, you are also helping to build the test to make Bots that can read or count or tell traffic lights from mountains or horses from humans. Seen lots of questions about driving lately? What could that be building a test for?
它可以工作，但这令人沮丧，尤其是当学生机器人非常擅长仅教给它的各种问题时。 照片非常棒，但是如果视频颠倒或显然不是蜜蜂的东西(例如猫)，则对视频无用或困惑不已。 我们的机器人机器人不能教所有东西，人类监督员所能做的就是给它更多的问题，使测试时间更长，甚至包括最好的机器人出错的问题。 了解这一点很重要。 这就是公司沉迷于收集数据的原因-更多数据等于更长的测试，等于更好的机器人。 当您在网站上进行人体测试时，不仅证明您是人类，而且希望您还可以帮助建立测试，以使Bot可以读取或计数或告诉山地或马匹的交通信号灯。人类。 最近有很多关于开车的问题吗？ 那可以为测试做些什么？
So everywhere on the internet, behind the scenes, there are tests to increase the user interaction, or set prices just right to maximize revenue or pick the pockets from all your friends. You like articles which your friends share. That’s one way of getting bots to understand what you will like, based on your contacts. If it’s testable, it’s teachable!
因此，在互联网上的每个地方，都在幕后进行了测试，以增加用户互动，或者设置价格恰到好处，以最大程度地增加收入或从所有朋友那里掏钱。 您喜欢朋友分享的文章。 这是使机器人根据您的联系人了解您想要的方式的一种。 如果可以测试，那就可以教学！
The teacher and the student bots will graduate from the warehouse to be the algorithm of its domain at least for a little while. We’re used to the idea that the tools we use even if we don’t understand them, someone does, but with our machines that learn we are increasingly in a position where we use tools or are used by tools that no one, not even their creators understand. We can only hope to guide them with the tests we make, and get comfortable with it as our algorithmic buddies are all around and not going anywhere.
老师和学生机器人将至少从仓库毕业一段时间成为其领域的算法。 我们已经习惯了这样的想法：即使我们不了解它们，我们也会使用所使用的工具，但是有人会使用。但是随着我们学习机器的发展，我们越来越处在我们使用工具或被任何人使用的工具所占据的位置。甚至他们的创作者都明白。 我们只能希望通过我们进行的测试来指导他们，并让他们感到满意，因为我们的算法伙伴无处不在。
Here, we used an example of bees and threes, just for convenience. Some real world applications include — early detection of cancer, where we train bots to understand various patterns of cancers that develop and to classify as to whether that cancer will be malignant (serious, urgent medical care), or benign (not so serious, minimal medical care); fixing airline seat prices, forecasting dangerous tornadoes months ahead of time, and the automatic defibrillators are just other examples of where machine learning is applied.
在这里，为方便起见，我们使用了蜜蜂和三只猫的示例。 现实世界中的一些应用包括- 癌症的早期检测 ，我们在其中训练机器人以了解正在发展的各种癌症模式，并对癌症是恶性(严重，紧急医疗)还是良性(不太严重，轻微)进行分类。医疗); 固定飞机座位价格，提前几个月预测危险的龙卷风以及自动除颤器只是应用机器学习的其他示例。
“Hey, will bots takeover the world?”
“嘿，机器人会占领世界吗？”
—Random Joe
-随机乔
“Worrying about bots taking over the world is just like worrying about overpopulation on Mars. We’re not even there yet!”
“担心机器人接管世界就像担心火星上的人口过多。 我们还没到那儿！”
— Andrew Ng, Machine learning Expert.
-机器学习专家Ng Andrew。
With that being said, I’d really like to recommend Andrew Ng’s course on Machine Learning and Deep Learning on Coursera, which I have personally benefited from.
话虽如此，我真的很想推荐Andrew Ng的 Coursera上有关机器学习 和 深度学习 的课程，我从中受益匪浅。
Also, I recommend the Kaggle platform, where you can learn machine learning skills for free, and test them too! They've got loads of contests every week and prizes up-to $10,000!
另外，我推荐Kaggle平台，您可以在其中免费学习机器学习技能 ，也可以对其进行测试！ 他们每周都有大量的比赛，奖品高达$ 10,000！
So to sum up, I’ll leave you with something that crossed my Social Media feed recently.
综上所述，我将为您提供一些最近跨过我的社交媒体供稿的内容。
Thanks for reading!
谢谢阅读！
If you liked this article, do check my other write-ups on Space Research, and also my first ever article on the A380 — The End of an Era.
如果您喜欢这篇文章，请查阅我在《 太空研究》 上的其他文章， 以及我在 A380 上发表的第一篇文章《 时代的终结》 。
Follow us for more amazing and latest science and technology articles. We are also there on social media. Follow us on Instagram|LinkedIn to stay updated!
关注我们，获取更多令人惊奇的最新科技文章。 我们也在社交媒体上。 在 Instagram上 关注我们 | 领英 保持更新！
HAVE A GREAT DAY!
祝你有美好的一天！
翻译自: https://medium.com/iete-sf-mec/how-do-machines-learn-an-algorithmic-approach-e8520fc0ba4b
机器学习是一种算法么
扫一扫
8万+
被折叠的 条评论
为什么被折叠?
到【灌水乐园】发言
