Augmenting Basin-Hopping With Techniques From Unsupervised Machine Learning: Applications in Spectroscopy and Ion...

Area of Science:

• Computational Chemistry
• Statistical Physics
• Molecular Biology
• Machine Learning

Background:

• Evolutionary algorithms like basin-hopping (BH) are valuable for complex optimization problems.
• BH maps potential energy surfaces (PES) to understand molecular dynamics and thermodynamics.
• Current methods can be computationally intensive for identifying local minima and transition states.

Purpose of the Study:

• To augment the basin-hopping (BH) algorithm with unsupervised machine learning techniques.
• To enhance the efficiency of identifying local minima and transition states on potential energy surfaces (PES).
• To demonstrate the utility of these integrated methods in interpreting experimental data.

Main Methods:

• Integration of unsupervised machine learning concepts: similarity indices, hierarchical clustering, and multidimensional scaling.
• Application to the basin-hopping (BH) algorithm for potential energy surface (PES) exploration.
• Utilizing machine learning for analysis of spectroscopic and ion mobility data.

Main Results:

• Successfully augmented the basin-hopping (BH) methodology with machine learning for efficient PES searches.
• Demonstrated the capability of machine learning techniques to rationalize experimental findings.
• Case studies confirmed the method's efficacy in spectral assignment and determining collision cross-sections.

Conclusions:

• Unsupervised machine learning significantly enhances basin-hopping (BH) algorithm performance in complex optimization.
• The integrated approach provides powerful tools for molecular modeling and experimental data interpretation.
• This synergy advances the study of molecular dynamics, thermodynamics, and related experimental investigations.