Experimenting a machine learning approach to geophysical rotation

Experimenting a machine learning approach to geophysical rotation

Earth -entering radar, described in use above, is a source of observation that uses geophysical rotation. Available: Comprehensive Nuclear-Test-Ban Treaty, CC BY 2.0

A common problem in geosciences is the need to distinguish between unidentified physical species based on limited information. For example, a radar observation that enters the earth attempts to compare the basement without in situ measurements. This type of problem is called transformation, in which the physical model is often corrected to bring it into harmony with the information.

The effects of the conversion can be greatly aggravated by the choice of features, which act as Bayesian in the first place. And because the features are more complex than the physical world, the process can result in an oversimplified result. To avoid these problems, it is common to add a model with what is seen in the real world, such as data collected from outcroppings or pit holes. This team can result in some modeling changes to give a real difference to the former.

Recent advances in this area have been made in the form of machine learning technologies. Convolutional neural systems such as those used in computer science have proven effective in combining multiple modeling models to create nuanced fronts while maximizing spatial resolution. Lopez-Alvis et al. see one such neural network approach: the variational autoencoder (VAE).

Various autoencoders can “restore” past training data. They can create new samples that are similar to, but not identical to, the features seen in the introductory images. The authors test this capability by comparing the studied VAEs using individual input images with the studied images in groups of images between the synthetic data and the real view.

An important result of the study was that VAEs trained using image collections performed better than those based on a single intervention. In fact, the combined VAE works almost like the best filter material for both synthetic and field data. Therefore, it is better to combine the training inputs into one VAE and perform one conversion.

This study was published in Journal for Geophysical research: Solid Earth.

When it comes to AI, can we leave out data?

More information:
J. Lopez – Alvis et al, Geophysical Inversion using a Variational Autoencoder to Model an Assembled Spatial Prior Uncertainty, Journal for Geophysical research: Solid Earth (2022). DOI: 10.1029 / 2021JB022581

Presented by the American Geophysical Union

This story was reprinted courtesy of Eos, maintained by the American Geophysical Union. Read the original story here.

Directions: Testing a way to teach a machine to geophysical inversion (2022, April 1) downloaded on 2 April 2022 from https://phys.org/news/2022-04-machine- approach-geophysical-inversion.html

This document is subject to copyright. Except for appropriate action for the purpose of personal inquiry or research, no piece may be reproduced without permission. Information is provided for informational purposes only.

Related Posts

Leave a Reply

Your email address will not be published.