IJUC Home • Issue Contents

Molecular Learning of a Soft-Disks Fluid
Luca Zammataro

This work is based on the Equivalence between Molecular Dynamics and Neural Network. It provides learning proofs in a Lennard-Jones (LJ) fluid, presented as a network of particles having non-bonded interactions. I describe the fluid’s learning as the property of an order that emerges as an adaptation in establishing equilibrium with energy and thermal conservation. The experimental section demonstrates the fluid can be trained with logic-gates patterns. The work goes beyond Molecular Computing’s application, explaining how this model uses its intrinsic minimizing properties in learning and predicting outputs. Finally, it gives hints for a theory on real chemistry’s computational universality.

Keywords: Molecular learning, molecular computing, molecular dynamics, thermodynamics, machine learning, Lennard-Jones fluid, self-organizing systems, logic-gates

Full Text (IP)