During my summer vacation, I ran across a CBInsights report called “AI trends to watch in 2020”. I was curious about what my colleagues and readers would think about the selected trends, so I launched a survey to see what they thought. I simply asked one question: “Based on your personal experience, which one is impacting your job/ industry the most?” and these were the results:

在暑假期间，我遇到了CBInsights的一份报告 ，该报告名为“ 2020年值得关注的 AI趋势”。 我很好奇我的同事和读者对所选趋势的看法，因此我发起了一项调查，以了解他们的想法。 我只是问了一个问题：“根据您的个人经验，哪一个对您的工作/行业影响最大？” 结果是：

The survey results were clear: AutoML was by far the trend that my audience considered to be impacting their job/ industry the most. In this article I will try to introduce you to the concept of AutoML and what can it mean for AI in the short term.

调查结果很清楚：到目前为止，AutoML是我的听众认为对他们的工作/行业影响最大的趋势。 在本文中，我将尝试向您介绍AutoML的概念及其对AI的短期意义。

什么是AutoML？ (What is AutoML?)

AutoML can be defined as a set of tools that can automate the process of solving problems with Machine Learning. Such a process involves several steps that require specific expertise in the field, such as data pre-processing, feature engineering, extraction, and selection. Not only that, Machine Learning experts must also select the right algorithm and perform optimization tasks in hyperparameters to maximize its accuracy.

可以将AutoML定义为一组工具，这些工具可以使机器学习解决问题的过程自动化。 这种过程涉及多个步骤，这些步骤需要该领域的特定专业知识，例如数据预处理，特征工程，提取和选择。 不仅如此，机器学习专家还必须选择正确的算法，并在超参数中执行优化任务，以使其准确性最大化。

When combined with MLOps methodologies and frameworks for at scale development and deployment of Machine Learning models, AutoML can become a interesting tool to democratize AI for business organizations.

当与MLOps方法和框架结合以进行大规模的机器学习模型开发和部署时，AutoML可以成为使商业组织的AI民主化的有趣工具。

为什么AutoML对于愿意采用AI的公司来说很有趣？ (Why is AutoML interesting for a company willing to adopt AI?)

AutoML can be a very good addition to regular Machine Learning projects for an organization that is willing to accelerate adoption. We can identify its goals and challenges organizing them if four categories, with the aim of achieving a Continuous Value Generation for AI initiatives:

对于愿意加速采用的组织，AutoML可以作为常规机器学习项目的很好的补充。 我们可以确定其目标和将其组织起来的挑战(如果有四个类别)，以实现AI计划的持续价值创造：

For each one of those areas, AutoML can help achieve goals related to AI democratization and scalable adoption. Nevertheless, it also presents some challenges for organizations, which are also listed below:

对于其中的每个领域，AutoML都可以帮助实现与AI民主化和可扩展采用有关的目标。 尽管如此，这也给组织带来了一些挑战，下面列出了这些挑战：

AutoML解决方案的当前前景如何？* (What is the current landscape of AutoML solutions?*)

First, it is important to note that you can find different kind of tools that can be categorized as AutoML. One way to classify them is by separating those that focus on specific parts of the Machine Learning process automation and those that try to cover the end to end. For the later, Forrester published its Forrester New Wave: Automation-Focused Machine Learning solutions in 2019, which concluded H20.ai and DataRobot are the most popular solutions for AutoML and, as Forrester mentions, are amongst the most mature ones.

首先，必须注意，您可以找到可以归类为AutoML的不同类型的工具，这一点很重要。 对它们进行分类的一种方法是，将那些专注于机器学习流程自动化特定部分的内容与那些试图涵盖端到端的内容分开。 后来，Forrester在2019年发布了其Forrester新浪潮：以自动化为中心的机器学习解决方案 ，并得出结论，H20.ai和DataRobot是AutoML最受欢迎的解决方案，并且正如Forrester所提到的，它们是最成熟的解决方案之一。

Although not mentioned on this Forrester report, I would also add Dataiku as one of the top commercial solutions.

尽管在此Forrester报告中未提及，但我还将Dataiku列为顶级商业解决方案之一。

DataRobot

数据机器人

DataRobot is probably the best known commercial solution for AutoML and one of the unicorns in the AI space, with a valuation above 1 billion dollars. In this blog post, they summarize the 10 top features that you should look for in an AutoML solution (and that, of course, DataRobot offers), which include:

DataRobot可能是AutoML的最著名商业解决方案，也是AI领域的独角兽之一 ，其估值超过10亿美元。 在此博客文章中 ，他们总结了您应该在AutoML解决方案中寻找的10个主要功能(当然，还有DataRobot提供的功能)，其中包括：

• Preprocessing of Data数据预处理
• Feature Engineering特征工程
• Diverse Algorithms多种算法
• Algorithm Selection算法选择
• Training and tuning训练和调优
• Ensembling组装
• Head to Head Model Competitions头对头模型竞赛
• Human-Friendly Insights人性化见解
• Easy Deployment易于部署
• Model Monitoring and Management模型监控与管理

DataRobot’s offering is composed of four independent products (Automated Machine Learning, Automated Time Series, MLOps and Paxata) that you can then consume in several ways, from a managed cloud to a private or hybrid cloud. You can find the detailed offering here.

DataRobot的产品由四个独立产品(自动化机器学习，自动化时间序列，MLOps和Paxata)组成，您可以通过多种方式使用它们，从托管云到私有云或混合云。 您可以在此处找到详细的产品 。

H2O.ai

过氧化氢

H20.ai defines it platform as “the open source leader in AI and machine learning with a mission to democratize AI for everyone”. The company offers an AutoML package as part of their Open Source platform, which includes the following features, according to their website:

H20.ai将其平台定义为“ AI和机器学习的开源领导者，其使命是使所有人都享有AI的民主化”。 该公司根据其网站提供AutoML软件包作为其开源平台的一部分，其中包括以下功能：

• Data preprocessing, including imputation, one-hot encoding, standardization. Feature selection and/ or extraction as well as count/ label/ target encoding of categorical features.数据预处理，包括插补，一键编码，标准化。 特征选择和/或提取以及分类特征的计数/标签/目标编码。
• Model generation, including Cartesian or random grid search and Bayesian hyperparameter optimization.模型生成，包括笛卡尔或随机网格搜索以及贝叶斯超参数优化。
• Ensembles.合奏。

H2O.ai also sells an enterprise version called Driverless AI, which they nicely summarize in an architectural diagram at their website.

H2O.ai还出售称为无人驾驶AI的企业版，他们很好地在其网站的架构图中进行了总结。

Dataiku

达蒂库

As the company mentions in one of their recent blog posts, not only they have focused in feature engineering AutoML capabilities for a long time, they also offer a visual tool that can train a model selecting the best models, features, etc. with only one button click (the user is still given the possibility to manually complete the different steps of the training process manually if he wants).

正如该公司在最近的一篇博客文章中提到的那样，他们不仅长期专注于功能工程AutoML功能，而且还提供了一种可视化工具，可以训练模型，仅用一个模型就能选择最佳模型，特征等。单击按钮(如果需要，用户仍然可以手动完成培训过程的不同步骤)。

Dataiku offers four product packagings, from free to enterprise. You can find how the different editions compare here.

Dataiku提供四种产品包装，从免费到企业。 您可以在这里找到不同版本的比较。

Besides these specific AutoML solutions, hyperscalers are starting to incorporate Automatic Machine Learning features to its development tools. The following is a non-exhaustive list of functionalities in which AI players like AWS, Google Cloud or Microsoft are starting to introduce AutoML capabilities.

除了这些特定的AutoML解决方案之外，超大规模开发者还开始将自动机器学习功能集成到其开发工具中。 以下是功能的详尽列表，其中AWS，Google Cloud或Microsoft等AI播放器开始引入AutoML功能。

Are you already using AutoML capabilities to scale and democratize AI? If so, I would love to hear about your experience in the comments section.

您是否已经在使用AutoML功能来扩展AI并使之民主化？ 如果是这样，我很想在评论部分听听您的经历。

*Please note that the information presented in this article was collected at the time of its preparation through the public websites of the different companies mentioned, so it may not be exact at the time of reading it.

*请注意，本文中提供的信息是在准备时通过提及的不同公司的公共网站收集的，因此在阅读时可能并不准确。