Statistical and machine learning models for the evaluation of geophysical and geomechanical data

This thesis concerns two critical issues of modelling in geophysics. Inverse problems are ubiquitous in nature. They are often used to study a geophysical phenomenon with more than one causative set of parameters. Their mathematical formulation involves expressing them using partial differential equations. When the system is computationally extensive, it can delay the generation of solutions and the interpretation process. We understand that supervised and dimensionality reduction algorithms can help resolve such issues. We develop deep learning models for two critical applications in geophysics. We show their efficacy in speeding up the synthetic data generation process and aiding automatic interpretation.