Google’s AI divi­sion re­leases open source up­date to Deep­Vari­ant

OpenSource For You - - Fossbytes -

Google has re­leased the open source ver­sion of Deep­Vari­ant, a deep learn­ing tech­nol­ogy to re­con­struct the true genome se­quence from HTS se­quencer data with sig­nif­i­cantly greater ac­cu­racy than previous clas­si­cal meth­ods. This work is the prod­uct of more than two years of re­search by the Google Brain team, in col­lab­o­ra­tion with Ver­ily Life Sciences.

The Brain team pro­grammed it in Ten­sorFlow, a li­brary of open source pro­gram­ming code for nu­mer­i­cal com­pu­ta­tion that is pop­u­lar for deep learn­ing ap­pli­ca­tions. The tech­nol­ogy works well on data from all types of se­quencers and eases the process of tran­si­tion­ing to new se­quencers.

Deep­Vari­ant is be­ing re­leased as open source soft­ware to en­cour­age col­lab­o­ra­tion and to ac­cel­er­ate the use of this tech­nol­ogy to solve real world prob­lems. The soft­ware will make it eas­ier to re­ceive in­puts from re­searchers about the use cases they were in­ter­ested in. This is part of a broader ef­fort to make the data of ge­nomics com­pat­i­ble with the way deep learn­ing ma­chin­ery works. Also, this move will push Google tech­nolo­gies to health­care and other sci­en­tific ap­pli­ca­tions, and make the re­sults of these ef­forts broadly ac­ces­si­ble.

Newspapers in English

Newspapers from India

© PressReader. All rights reserved.