|
所在平台: Udemy |
课程主页: https://www.udemy.com/course/explainableai/
课程评论：没有评论
Coursera 课程：可解释人工智能——解开 AI 模型“黑箱” **课程简介：** 本课程专为希望理解和掌握人工智能模型（特别是机器学习模型）内部运作机制的学员设计。课程以一个实际场景开篇：一位机器学习工程师面对非技术团队关于欺诈检测模型为何将某笔交易标记为欺诈的质疑。模型如同一个“黑箱”，工程师难以给出清晰合理的解释。 **课程目标：** * **拆解“黑箱”：** 教授学员如何理解和解释 AI 模型的预测结果，使其透明化、可信赖。 * **理解可解释性的重要性：** 阐述为何可解释性是部署 AI 的关键要素。 * **掌握解释方法：** 深入讲解局部解释（individual predictions）和全局解释（overall logic of models），以及反事实解释（counterfactuals）的概念——即改变哪些因素可以导致模型做出不同决策。 * **学习 SHAP 库：** 重点介绍 SHAP (SHapley Additive exPlanations) 这一强大的工具库，用于揭示特征对模型预测的贡献度。 * **提升沟通能力：** 使学员能够将抽象的 AI 结果转化为易于理解和令人信服的解释。 **核心内容：** * AI 模型透明化的必要性。 * 局部和全局解释技术。 * 反事实推理及其应用。 * SHAP 库的使用与原理。 **课程价值：** 通过本课程，学员将能够自信地解释 AI 模型的决策过程，从而提升 AI 在实际应用中的可信度、问责制和可理解性，使 AI 成为一个强大的、可控的工具。
Imagine a scenario where a Machine Learning Engineer, armed with a sophisticated fraud detection model, is struggling to justify its outcomes to a non-technical team. Questions are fired from all corners:"Why was this transaction flagged as fraudulent?""What factors led to this decision?""Can we trust these results?" The ML engineer is at a loss - the model is a black box, and deciphering it seems like an enigma. If you've ever found yourself in a similar situation or have asked these kind of questions to your machine learning team, our course, "Explainable AI", is tailor-made for you.We believe in teaching without detours, getting straight to the point and without beating around the bush. Our aim? To equip you with the skills to crack open the 'black box' of AI, making it transparent and trustworthy.We illuminate the realm of explainability, dissecting why it's a cornerstone for any AI deployment. With a focus on both local and global explainability, we demonstrate how to dissect individual predictions and unravel the overall logic of models. The intriguing concept of counterfactuals is explored, painting a picture of alternative scenarios that could alter a model's decision.We also dive deep into the world of SHAP (SHapley Additive exPlanations), an invaluable library that unearths the contributions of features in model predictions. By the end of this course, you'll be able to transform abstract AI outcomes into understandable, convincing explanations. So, let's together demystify AI and ensure it becomes an accountable and comprehensible tool in your arsenal!
