Machine Learning in Geomechanics 1

3Unsupervised Learning: Basic Concepts and Application to Particle Dynamics

Noel JAKSE

SIMaP, CNRS, Grenoble INP, Université Grenoble Alpes, France

3.1. Introduction

Machine learning (ML) approaches attracted significant interest in many scientific fields due to their potential ability to uncover patterns, make predictions and extract valuable insights from large and complex body of data (Efron and Hastie 2021). Web-based data collection platforms, scientific instruments and computer simulations are creating exponentially increasing data stores. They triggered new scientific methods to analyze and organize huge amount of data, giving the possibility to find subtle effects missed previously (Hey et al. 2009; Hey and Trefethen 2020).

ML tools can be broadly categorized into three types: supervised, unsupervised and semi-supervised learning. Each type offers distinct approaches to learning from data, with specific strengths and limitations (Efron and Hastie 2021; Hastie et al. 2009). Supervised learning is one of the most used approachs and involves training a model using labeled data, where both input features and corresponding output labels are provided. The model learns to map inputs to their respective outputs, generalizing to make predictions for new, unseen data. Applications of supervised learning include regression and classification tasks that will be the subject of Chapters 2 and 4.

However, obtaining labeled data can be time consuming and expensive, especially for ...