课程: Foundations of Responsible AI

今天就学习课程吧！

今天就开通帐号，24,700 门业界名师课程任您挑！

Building data understanding

Building data understanding

课程: Foundations of Responsible AI

开始一个月试用 购买团队方案

Building data understanding

“

- Let's flick some of our bias identification skills by looking at historical housing data, despite our goals it's unfortunately easy to perpetuate bias with our algorithms. Let's remember for a moment that the purpose of machine learning models are to extract some kind of understanding of why past decisions were made and apply that logic to new data. Housing data represents actual housing decisions like who is approved for a home loan and the humans involved in making those decisions may have been discriminatory. Thinking critically about what could go wrong when using this data is crucial because it's simple to perpetuate biases with our algorithms. In this example if we investigate just one protected class, race, we can see that systemically homes in Black neighborhoods are sold for less and appreciate in value slower than homes in mostly white neighborhoods. If we train a model that's 90% accurate on this data we'll…

内容