Statistical learning and regularisation for regression: Application to system identification and time series modelling

Doctoral Thesis / Dissertation from the year 1997 in the subject Computer Sciences - Artificial Intelligence, grade: n/a, language: English, abstract: Unravel the complexities of data-driven modeling and step into the realm where accuracy meets resilience in the face of uncertainty. This groundbreaking work delves deep into the heart of statistical learning and regularization, offering a transformative perspective on regression techniques. Discover how these powerful methods can revolutionize system identification and time series modeling, unlocking unprecedented levels of precision and reliability. Explore cutting-edge approaches to model parameter estimation, fortified by rigorous mathematical frameworks and innovative algorithms designed to combat overfitting and enhance generalization. Witness the evolution of model robustness as regularization techniques are meticulously applied to mitigate the impact of noisy data and spurious correlations. This research bridges the gap between theoretical advancements and practical applications, providing invaluable insights for researchers and practitioners alike. Whether you're grappling with the intricacies of dynamic systems or seeking to forecast future trends, this exploration of statistical learning, regularization, and regression will equip you with the tools and knowledge to conquer even the most challenging modeling tasks. Journey through the landscape of data analysis, guided by a comprehensive treatment of both established techniques and novel methodologies, and emerge with a profound understanding of how to build models that are not only accurate but also resilient and insightful. From the foundations of regression to the frontiers of system identification and time series analysis, this is your compass for navigating the ever-evolving world of data-driven decision-making. Prepare to challenge conventional wisdom, embrace the power of regularization, and unlock the true potential of your data. The exploration of