Reinforcement Learning Based Resolution Improvement of Geophysical Data
- Tech Stack:Python, matplotlib, seaborn, reinforecement learning
- Report Link: Project Link
- Github Link: Project Link
Majority of magnetic field of the Earth originate from outer core region. Models are required to observe the variation and study the origin of magnetic field all over the globe. Also, magnetic field changes abruptly through time and location on surface. Prior to advancement of satellite measurement, we have limited observed data. Our aim to develop a model which has ability to predict field value in the past and also to have a better resolution. In order to develop model, evolution is required to have more accuracy. But due to lack of data it is hard to made accurate model for field data prediction.
Hence our approach is to use the Reinforce learning along with physics behind its nature of variation. We require some paleomagnetic criteria to further evaluate our model. In this work we have tried to find an approach to use reinforcement learning technique which might be helpful to design a model that can predict earth’s magnetic field in past events with high resolution. Also, we have tried to analyze the paleomagnetic criteria and measure the fit level with the current globally accepted model IGRF-13.