![“How did software get so reliable without proof?” is the title of a classic article by C. A. R. Hoare [1]. He asked this question already in 1996.](http://static1.squarespace.com/static/60886120e95d27168bb0eea5/t/60e41a6794b0d322fb1446b3/1625561708276/IMG_6153.jpg?format=1500w "Test machine learning the right way: regression testing")
Test machine learning the right way: regression testing
“How did software get so reliable without proof?” is the title of a classic article by C. A. R. Hoare [1]. He asked this question already in 1996. And really, when was the last time you entered a search term on Google and got no results? The answer for me is never. I can’t even remember the last time the search results weren’t relevant. So, how did software get so reliable? The simple answer is – through rigorous development processes, continuous improvement of existing software, and extensive testing.
This leads us to the question, when, if ever, will we be able to ask the same question about machine learning (ML)? I’m optimistic, however, it won’t happen with the current level of testing methodology and test tools specifically aimed at machine learning!
In this blog series, we’ll investigate how we can better test machine learning applications to improve their reliability and increase performance. In the first post, we’ll look at what we mean by ML testing, what an ML bug is, and where they occur, as well as introduce the first technique for your ML testing repertoire: regression testing.
