For those that went to the MoT Cambridge meetup on Thursday September 12, 2019, slides are attached.
Apologies if this is not the best place to share. I can move or remove if necessary.
Talk Summary:
JOHN MCLEOD: ARTIFICIAL INTELLIGENCE - TESTING MACHINE LEARNING CODE
Can you write tests of machine learning code? This is an important question as machine learning models and algorithms are turning up in AI based software that is becoming part of many products and services.
CAN you write tests of machine learning code? Yes! In this talk I will introduce some aspects of machine learning (specifically reinforcement learning), and give some examples of the sorts of tests that can be written.
Can YOU write tests of machine learning code? Yes! Lots of machine learning code is written in a way that makes it hard to test. In addition, it is often difficult to define what “working correctly” really means. Fortunately, there is usually another way to approach the testing problem which can ensure the quality of the code and the correctness of the implementation.
Original meetup link:
Testing Machine Learning Code.pdf (568.8 KB)