In recent years, a wealth of tools has appeared that automate the machine learning cycle inside a black box. At KNIME we take a different stance. Automation should not result in black boxes, as modern data science should allow automation and interaction to be combined flexibly into a more transparent solution. Paolo will show how to build such a more transparent application for automated machine learning using KNIME Software. It will have an input user interface to control the settings for data preparation, model training (e.g. using deep learning, random forest, etc.), hyperparameter optimization, and feature engineering. The trained models will be shown in the end in an interactive dashboard to visualize the results with model interpretability techniques.
Bio: Paolo Tamagnini works as a Data Scientist for KNIME in the evangelism team in Berlin. After graduating with a Master’s Degree in Data Science at Sapienza University of Rome, Italy, Paolo gathered research experience at New York University in machine learning interpretability and visual analytics tools. Since working for KNIME, Paolo has run various workshops in the US and Europe and has developed a number of reusable guided analytics applications for automated machine learning and data exploration.