Researchers From China Proposed A Novel μXRD Image Screening Method Based On Federated Learning (FL) To Improve The Screening While Protecting Data Privacy

Federated learning provides a mechanism to unleash data to fuel new AI applications by training AI models without letting anyone see or access your data. Industrial minerals are subjected to synchrotron X-ray microdiffraction (XRD) services to determine their crystal impurities in terms of crystallinity and possible impurities. XRD services produce huge amounts of photos; these images must be filtered before being processed and stored further. Due to service users’ reluctance to provide their original experimental photographs, there aren’t enough efficient labeled examples to train a screening model. To enhance screening while maintaining data privacy, researchers suggest federated learning (FL) based XRD image screening approach in this study. With the help of cutting-edge client sampling algorithms, their solution addresses the problem of imbalanced data distribution faced by service users while using various types and quantities of samples. They also suggest hybrid training techniques to address asynchronous data exchanges between FL clients and servers. The results of the studies demonstrate that their technology may guarantee efficient screening for commercial customers testing industrial materials while protecting commercially sensitive information.

Industrial minerals can detect crystal imperfections using synchrotron X-ray microdiffraction (XRD). However, the advancement of precise XRD image screening is being hampered by two significant issues. One is the dearth of labeled industrial samples, and the other is the industrial XRD service users’ privacy concerns.

The researchers used the physical information specific to the domain to increase the accuracy of federated learning. They then implemented a sampling method with new client sampling algorithms after considering the uneven data distributions in the real world. A hybrid training architecture was developed to deal with the erratic communication environment between FL clients and servers.

Extensive testing revealed that sharing data characteristics among users or applications without compromising commercially sensitive information increased the accuracy of machine learning models by 14% to 25%. This innovative system’s federated learning capabilities will assist remove non-technical obstacles to data exchange.

This Article is written as a research summary article by Marktechpost Staff based on the research paper  'Federated Learning-Based Synchrotron X-ray Microdiffraction Image Screening for Industry Materials'.  All Credit For This Research Goes To Researchers on This Project. Check out the paper and reference article.

Please Don't Forget To Join Our ML Subreddit