July 13, 2019 09:03 EDT
There are several competitive theories about some phenomena in both quantum physics. But do both of them best describe quantum phenomena? A team of researchers from Munich Technical University (TUM) and Harvard University (USA) have successfully deployed artificial neural networks for image analysis of quantum systems.
Is that a dog or a cat? This classification is a representative example of machine learning. Artificial neural networks can be trained to find patterns that are characteristic of particular objects and to analyze images. If the system has acquired such a pattern, any figure can recognize a dog or cat.
Using the same principle, neural networks can detect tissue changes in the radiation image. Physicists now analyze the image of a bilateral multi-body system (a so-called snapshot) and find a theory that best describes observed phenomena.
Quantum World of Probability
Many phenomena in the physics of condensed matter physics that study solids and liquids are still puzzled. For example, it is not yet clear why the electrical resistance of a high-temperature superconductor drops to zero at a temperature of about -200 degrees Celsius.
Understanding these abnormal conditions is not easy. Quantum simulators based on cryogenic lithium atoms have been developed to study the physics of high-temperature superconductors. They take snapshots of quantum systems that exist simultaneously in different configurations. Physicists talk about nesting. Each snapshot of a quantum system provides one specific configuration depending on the quantum mechanical probability.
Various theoretical models have been developed to understand these quantum systems. But how well do you reflect reality? Questions can be answered by analyzing image data.
Neural networks study the quantum world.
To do this, a team of researchers from the University of Munich and Harvard University successfully conducted machine learning. The researchers trained artificial neural networks to distinguish the two competition theories.
Annabelle Bohrdt, a Ph.D. student at TUM, says, "Just like cat or dog detection in a picture, the composition image of every quantum theory enters the neural network." Network parameters are optimized to give each image the correct label. A cat or dog, not theory A or theory B. "
After completing the training phase with the theoretical data, the neural network had to apply the learning and assign a snapshot of the quantum simulator to the theory A or B. So the network has chosen a more predictable theory.
We plan to use this new method in the future to evaluate the accuracy of multiple theoretical explanations. The purpose of this study is to understand the main physical effects of high-temperature superconductivity in which there are many important applications, as well as two examples of lossless power transmission and efficient magnetic resonance imaging.
© 2018 NatureWorldNews.com All rights reserved. Do not reproduce without permission.