At the conclusion of 2021, researchers at UNC Charlotte and Duke University Medical Center published results of work done to use transfer learning methods to generate fluence maps for radiation therapy, aimed at providing medical professionals with more capability and information in fighting adrenal cancers. The technique uses a deep transfer learning model trained on a much larger dataset that can be applied to a smaller data set for a specific application.
The initial model was trained on pancreas treatment plans, then retuned and applied to a smaller set of data points on adrenal cancers. The output of the model generates a fluence map for specific IMRT beam-based treatments for adrenal cancers.
According to the researchers, this approach is meant to supplement but not replace human expertise in the field and is reliant on human expertise to finetune and improve the AI model.
European Union Seeking to Strengthen Semiconductor Ecosystem
Mark Nossokoff, Bob Sorensen
On February 8, 2022, the European Commission formally proposed what's commonly referred to as the European Chips Act. The legislation plans to build on Europe's strengths and address weaknesses to develop a thriving domestic semiconductor ecosystem and resilient supply chain, while setting measures to anticipate and respond to future supply chain disruptions. In the short term, the Act seeks to bolster EU capabilities to anticipate future chips crises, strengthen manufacturing activities in the EU, and support scale-up and innovation across the whole value chain. In the mid- to long-term, it seeks to reinforce Europe's technological leadership while developing mechanisms to support transfer of knowledge from the lab to the fab and position Europe as a technology leader in innovative downstream markets.
2 202022 | HYP_Link
Barcelona Supercomputing Center Trains Spanish NLP Model
Alex Norton, Bob Sorensen
Recently, the Barcelona Supercomputing Center (BSC) trained the first large artificial intelligence (AI) model designed to understand, speak, and write in the Spanish language. The system, named MarIA, was trained on the MareNostrum supercomputer at the BSC, leveraging 59 TBs of language data from the Biblioteca Nacional de España, one of the world's largest public libraries. The model is said to be an expert in both writing and understanding the Spanish language and is free to use by any developer, company, or entity. The system has a wide variety of potential applications including summarization applications, chatbots, smart searches, translation engines, and automatic subtitling chatbots.