Machine Learning (ML) techniques are becoming more accessible to develop with the availability of open source software, an increase of data available across industries, and accessibility of online learning platforms to upskill people. Many industries are starting to adopt these techniques as a way to solve business problems, automate processes, improve profits or reduce costs. But what happens when a company wants to solve a problem through a ML model and simply it does not work. Where did the model go wrong?
In this talk, Eduardo will go through the process of developing a ML model. At the different model development stages, he will highlight the components where a model can introduce biases and errors and will describe what are the techniques that can help to correct them. Once the model is developed, he will go through methods to interpret the ML models and how to verify if the model is producing fair outcomes.
Technical level: Technical practioner