Browsing by Author "Ferrara, Massimiliano"

Now showing 1 - 2 of 2

Optimal Control in Geometric Dynamics: Applications To AI Algorithm Optimization
(Universal Wiser Publisher, 2025) Ferrara, Massimiliano
This paper extends the fundamental concepts of geometric dynamics and optimal control theory, inspired by the pioneering work of Professor Constantin Udriste, to develop novel optimization algorithms for artificial intelligence systems. We establish connections between nonholonomic macroeconomic systems and reinforcement learning by formulating a multi-time maximum principle framework that integrates sub-Riemannian geometry. Our proposed methodology demonstrates how constrained variational problems can optimize neural network training trajectories through a bang-bang control approach. An empirical case study implements this theoretical framework to optimize a deep reinforcement learning algorithm, showing significant improvements in convergence speed and stability compared to standard approaches. The results demonstrate the practical value of geometric dynamics principles in modern Artificial Intelligence (AI) optimization, establishing a bridge between classical mathematical control theory and contemporary machine learning challenges.
Solow-Swan Model and Growth Dynamics: Moving Forward
(Springer Int Publ Ag, 2025) Ferrara, Massimiliano
This note in the Milestones series is dedicated to the Solow-Swan model. The aim is to examine the historical significance and enduring impact of the Solow-Swan neoclassical growth model, independently developed by Robert Solow and Trevor Swan in 1956. The model revolutionized economic growth theory by introducing a framework explaining long-term growth through capital accumulation, labor growth, and technological progress. We explore the model's theoretical foundations, influence on subsequent literature, empirical applications, and ongoing relevance. The paper presents novel extensions with discrete time delays that provide insights into cyclical economic phenomena, demonstrating how time-to-build technology can generate endogenous fluctuations within the otherwise stable Solow framework.