Tutorials

Pavel Sherstnev. Thefittest Library: Evolutionary Algorithms and Automation of Machine Learning Models Design in Python

Vladimir Stanovov. Modern differential evolution: from simple adaptive schemes to hyper-heuristics

Thefittest Library: Evolutionary Algorithms and Automation of Machine Learning Models Design in Python

Evolutionary algorithms are powerful tools for solving a wide range of optimization tasks. This tutorial introduces Thefittest — a flexible Python library that provides access to modern evolutionary modelling and optimization methods for both machine learning model design and general optimization tasks.

We will begin with a brief overview of evolutionary algorithms — their benefits and main application areas — then take an in-depth look at Thefittest, its architecture, and unique features. Using practical examples, we will demonstrate how the library applies to tasks like model building and optimization, highlight its advantages, and compare it with other widely used tools in evolutionary computing.

Modern differential evolution: from simple adaptive schemes to hyper-heuristics

Differential evolution has shown itself to be one of the best heuristic numerical optimizers, according to many competitions in the last decade. Its simplicity and efficiency have allowed it to become very popular among researchers and practitioners in many fields.

In this tutorial we shall consider its history, starting from the simplest variants, which can be easily implemented in any programming language, up to current state-of-the-art approaches, such as L-SHADE. The focus of this tutorial will be on the adaptive differential evolution, and how it may help solving practical problems. In addition, several findings using of the hyper-heuristic approach will be considered, which allowed creating the best numerical optimizers.