反事实 机器学习_关于机器学习的5个不可否认的事实

反事实 机器学习

Today, devices are trained to replicate human intelligence. While the model of machines automating various tasks have gained acceptance and is evolving, it comes up with a number of facts and notions. Similar is the case with Machine Learning. Here we are with some of the verities about ML that I think everyone should be aware of.

如今，已经对设备进行了训练以复制人类智能。 虽然使各种任务自动化的机器模型已经获得接受并且正在发展，但它提出了许多事实和观念。 机器学习也是如此。 我认为每个人都应该意识到有关ML的一些事实。

1. Machine Learning is Different from AI or Data Mining

1.机器学习不同于AI或数据挖掘

Artificial Intelligence (AI) is an umbrella term, which is given to computer systems that perform tasks, requiring human efforts. Very often, the terms AI, ML, and data mining are used interchangeably, but all of them are completely different. Let’s understand why.

人工智能（AI）是一个笼统的术语，被赋予执行任务并需要人工的计算机系统。 通常，术语AI，ML和数据挖掘可以互换使用，但它们完全不同。 让我们了解为什么。

AI is a machine programmed to think like a human. If we talk about human the brain, it could be described as one of the finest computing machines in existence. At any given time, it can capture tons of data. Using it's five senses, it saves, recalls, and processes what it's captured whenever needed to make informed decisions. It learns by recognizing patterns and this is one of the most effective examples of cognitive learning. But, as they say, even geniuses have a limit, and so does the human brain.

人工智能是一种被编程为像人一样思考的机器。 如果我们谈论人类的大脑，它可以说是现有的最好的计算机之一。 在任何给定时间，它都可以捕获大量数据。 使用五种感觉，它可以保存，调用和处理在需要进行明智决策时所捕获的内容。 它通过识别模式进行学习，这是认知学习最有效的例子之一。 但是，正如他们所说，即使是天才也有局限性，人脑也是如此。

So, you can think of machine learning as an automated and continuous version of data mining. And these data sets can be of any size. ML can analyze dynamically changing big data sets, detect and extrapolate the patterns, derive information, and apply it to new solutions and actions.

因此，您可以将机器学习视为数据挖掘的自动化和连续版本。 这些数据集可以是任意大小。 ML可以分析动态变化的大数据集，检测和推断模式，导出信息并将其应用于新的解决方案和行动。

2. Machine Learning means Data and Algorithms Together

2.机器学习将数据和算法结合在一起

Machine Learning is enabling the computers to learn, without being explicitly programmed for the same. The entire idea is based on cognitive learning, wherein the former actions are analyzed to process the recent input. This means, for learning to give an output, data is the key. The more data, the better the experience and accurate the result will be. 

机器学习使计算机能够学习，而无需对其进行显式编程。 整个构想是基于认知学习的，其中分析了先前的动作以处理最近的输入。 这意味着，对于学习给出输出，数据是关键。 数据越多，体验越好，结果将更加准确。

There’s a lot of excitement about advances in machine learning algorithms, and particularly about deep learning. But data is the key ingredient that makes machine learning possible. You can have machine learning without sophisticated algorithms, but not good data.

关于机器学习算法的进步，尤其是关于深度学习，有很多令人兴奋的事情。 但是数据是使机器学习成为可能的关键因素。 您可以在没有复杂算法的情况下进行机器学习，但没有好的数据。

3. Machine Learning is Incomplete without Humans

3.没有人，机器学习是不完整的

Undoubtedly, machine learning is adding more power to how humans perform their day to day tasks. However, that should not be mistaken for the possibility that machines will eventually do away with the need for human intelligence.

毫无疑问，机器学习正在为人类如何执行日常任务增加更多的力量。 但是，不要误以为机器最终会消除对人类智能的需求。

Even though machines will learn and will be smart enough, they still require human operators to build ML models, provide context, set parameters of operation, and do the needful to augment the algorithm.  

尽管机器将学习并且足够聪明，但是它们仍然需要操作员构建ML模型，提供上下文，设置操作参数以及进行必要的扩展算法。

ML is the competency of machines to recognize the complex patterns that humans can’t. Nevertheless, machines are clueless about the fact as to why those patterns exist. It’s the human who decides how the machines should work.

ML是机器识别人类无法识别的复杂模式的能力。 尽管如此，机器对于这些模式为何存在这一事实一无所知。 决定机器如何工作的是人。

4. Machine Learning is Vulnerable to Human Error

4.机器学习易受人为错误的影响

If the machine learning fails, it’s because of an algorithm failure. No matter how relevant or qualitative the data is, a human error in generating patterns and designing algorithms can lead to unbiased error, which ultimately ruins the system. Therefore, the best practice is to approach ML with a discipline and update it regularly for improved output and error detection.

如果机器学习失败，那是因为算法失败。 无论数据的相关性或质量如何，在生成模式和设计算法时出现人为错误都会导致无偏错误，最终会破坏系统。 因此，最佳实践是按照学科来处理ML，并定期对其进行更新，以改善输出和错误检测。

5. Machine Learning Follows Garbage In Garbage Out

5.机器学习跟进垃圾

What you give is what you get is the universal law. While it defines that relevant data and algorithm is the key to ML performance, it also defines its limitation. ML is only capable of responding to the data and patterns that are predefined. For supervised machine learning, it is important to categorize and train that data well.

你付出的就是普遍法则。 它定义了相关数据和算法是机器学习性能的关键，同时也定义了它的局限性。 ML仅能够响应预定义的数据和模式。 对于有监督的机器学习，正确地分类和训练数据很重要。

Certainly, Machine Learning brings in a number of advantages for every industry. We have seen how chatbots have been of huge importance in HealthIT, eCommerce, Sales, and many other segments. For those reasons, AI Application Development is having a huge demand in the business world. 

当然，机器学习为每个行业带来许多优势。 我们已经看到了聊天机器人如何在HealthIT，电子商务，销售和许多其他领域中发挥着巨大的作用。 由于这些原因，AI应用程序开发在商业世界中有着巨大的需求。

翻译自: https://www.experts-exchange.com/articles/31302/5-UnDeniable-Facts-about-Machine-Learning.html

反事实 机器学习