登录查看更多内容

What are the best ways to validate machine learning and artificial intelligence assumptions?

由人工智能和领英社区提供技术支持

Machine learning and artificial intelligence (ML/AI) are powerful tools for solving complex problems, but they also require careful validation to ensure that they are reliable, accurate, and ethical. Validation is the process of checking whether the assumptions and hypotheses that underlie ML/AI models and systems are true and consistent with the data and the real world. In this article, you will learn some of the best ways to validate your ML/AI assumptions, from data exploration and preprocessing to testing and evaluation.

在这篇协作文章中查找专家回答

由社区从 3 条内容中精选。了解更多

1 Data exploration

Before you build any ML/AI model or system, you need to understand the data that you are working with. Data exploration is the first step of validation, where you perform descriptive and visual analysis of the data, such as summary statistics, distributions, correlations, outliers, and missing values. Data exploration helps you to identify the characteristics, quality, and limitations of the data, as well as the potential features and targets for your ML/AI model or system.

添加您的观点

Maximilian Jalea

Machine Learning Engineer
举报内容
You ?need to know“ your data inside out. Important to look for especially after getting a first general glimpse are edge cases and outliers. Depending on your data, this can be a tricky part, as it may not easily be defined what your edge case might be.

已翻译

赞

2 Data preprocessing

After you have explored the data, you need to prepare it for ML/AI modeling or system development. Data preprocessing is the second step of validation, where you apply various techniques to clean, transform, and enhance the data, such as imputation, normalization, encoding, scaling, and feature engineering. Data preprocessing helps you to improve the performance, accuracy, and robustness of your ML/AI model or system, as well as to reduce noise, bias, and complexity.

添加您的观点

Maximilian Jalea

Machine Learning Engineer
举报内容
Don’t forget that it is important to understand what you are doing with your data. A first step can be to look again into your data, see what changed and how. The goal in the end is to not only get a deep understanding of your data, but also after it completed the preprocessing. Where can be pitfalls? What worked on other data already and where did it fail? Does the preprocessing do what you think it does?

已翻译

赞

3 Model selection

Once you have preprocessed the data, you need to choose the best ML/AI model or system for your problem. Model selection is the third step of validation, where you compare and evaluate different ML/AI algorithms, architectures, and parameters, such as linear regression, neural networks, decision trees, and hyperparameters. Model selection helps you to find the optimal trade-off between complexity and generalization, as well as to avoid overfitting and underfitting.

添加您的观点

Maximilian Jalea

Machine Learning Engineer
举报内容
A interesting question in the end is not only, how good the models/AI works, but also how much compute it needs and therefore how much time it takes or money it costs. Having these questions in mind already in the selection phase can help to get a wider field of possibilities.

已翻译

赞

4 Model testing

After you have selected the best ML/AI model or system, you need to test its performance and accuracy on unseen data. Model testing is the fourth step of validation, where you split the data into training, validation, and testing sets, and use various metrics and methods to assess the ML/AI model or system, such as accuracy, precision, recall, F1-score, ROC curve, and cross-validation. Model testing helps you to measure how well your ML/AI model or system can generalize to new data and scenarios, as well as to identify and correct errors and weaknesses.

添加您的观点

5 Model evaluation

Once you have tested the performance and accuracy of your ML/AI model or system, you need to evaluate its impact and implications on the real world. Model evaluation is the fifth step of validation, where you examine the ethical, social, and legal aspects of your ML/AI model or system, such as fairness, transparency, accountability, privacy, and security. Model evaluation helps you to ensure that your ML/AI model or system is aligned with the values and expectations of the stakeholders and the society, as well as to mitigate potential risks and harms.

添加您的观点

6 Model improvement

After you have evaluated the impact and implications of your ML/AI model or system, you need to improve its quality and value. Model improvement is the sixth and final step of validation, where you use feedback, insights, and best practices to refine and optimize your ML/AI model or system, such as feature selection, parameter tuning, regularization, ensemble methods, and deployment strategies. Model improvement helps you to enhance the reliability, efficiency, and usability of your ML/AI model or system, as well as to adapt to changing data and environments.

添加您的观点

7 Here’s what else to consider

This is a space to share examples, stories, or insights that don’t fit into any of the previous sections. What else would you like to add?

添加您的观点

Computer Science

+ 关注

给文章评分

我们借助人工智能创建了此文章。您认为这篇文章怎么样？

很棒不太好

举报此文章

查看全部

What are the best ways to validate machine learning and artificial intelligence assumptions?

1

2

3

4

5

6

7

1 Data exploration

2 Data preprocessing

3 Model selection

4 Model testing

5 Model evaluation

6 Model improvement

7 Here’s what else to consider

Computer Science

给文章评分

感谢您的反馈

更多Computer Science相关文章

更多相关阅读内容

What are the best ways to validate machine learning and artificial intelligence assumptions?

1

2

3

4

5

6

7

1 Data exploration

2 Data preprocessing

3 Model selection

4 Model testing

5 Model evaluation

6 Model improvement

7 Here’s what else to consider

Computer Science

给文章评分

感谢您的反馈

查看其他技能