MLFlow: Easily Optimize your Machine Learning Experiment

MLFlow at least attempts to turn your chaos into slightly more organized chaos

Introduction

Imagine, you experiment with some data. Regardless if it is text processing, image recognition, or a simple linear regression task, you will perform a sequence of data preprocessing modifications, test several models with different hyperparameters, and then measure its metrics. This already sounds chaotic, and most likely, if you just go with a brute-force notebook without any other tools, your experiment will turn into a mess with a lot of data and models randomly saved in the middle of your document. Such a way is barely efficient and hardly enjoyable. This is the moment when MLFlow comes in handy.

Diagram for hyperparameters tuning created with mlflow

MLFlow is an open-source framework for making your machine-learning experiment simple and efficient. It will save your hyperparameters, metrics, and models in one place and depict them graphically (like in the picture above) so that you can easily analyze the results of your experiment. There is no point in describing it more, so let’s learn by doing!

Task description

For this project I decided to implement something simple to focus not on the data but on the process of writing code. From school, we all know the task of finding the area under…