Esses algoritmos de IA exigiriam apenas uma unidade de processamento gra¡fico (GPU) de baixo custo, como as encontradas em sistemas de videogame, para processar dados LIGO avana§ados mais rápido do que em tempo real.
![](data:image/jpeg;base64,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)
Visualização cientafica de uma simulação da relatividade numanãrica que descreve a colisão de dois buracos negros consistente com a fusão do buraco negro bina¡rio GW170814. A simulação foi feita no supercomputador Theta usando o software de ca³digo aberto, relatividade numanãrica, Einstein Toolkit (https://einsteintoolkit.org/). Crédito: Argonne Leadership Computing Facility, Visualization and Data Analytics Group [Janet Knowles, Joseph Insley, Victor Mateevitsi, Silvio Rizzi].)
Quando as ondas gravitacionais foram detectadas pela primeira vez em 2015 pelo avana§ado Laser Interferometer Gravitational-Wave Observatory (LIGO), elas enviaram uma onda pela comunidade cientafica, pois confirmaram outra das teorias de Einstein e marcaram o nascimento da astronomia das ondas gravitacionais. Cinco anos depois, várias fontes de ondas gravitacionais foram detectadas, incluindo a primeira observação de duas estrelas de naªutrons colidindo em ondas gravitacionais e eletromagnanãticas.
Amedida que o LIGO e seus parceiros internacionais continuam a aprimorar a sensibilidade de seus detectores a s ondas gravitacionais , eles sera£o capazes de sondar um volume maior do universo, tornando a detecção de fontes de ondas gravitacionais uma ocorraªncia dia¡ria. Este dilaºvio de descobertas lana§ara¡ a era da astronomia de precisão que leva em consideração os fena´menos mensageiros extra-solares, incluindo radiação eletromagnanãtica, ondas gravitacionais, neutrinos e raios ca³smicos. A realização desse objetivo, no entanto, exigira¡ um repensar radical dos manãtodos existentes usados ​​para pesquisar e encontrar ondas gravitacionais.
Recentemente, o cientista computacional e lider de inteligaªncia artificial translacional (IA) Eliu Huerta do Laborata³rio Nacional de Argonne do Departamento de Energia dos EUA (DOE), em conjunto com colaboradores de Argonne, da Universidade de Chicago, da Universidade de Illinois em Urbana-Champaign, NVIDIA e a IBM, desenvolveu uma nova estrutura de IA em escala de produção que permite a detecção acelerada, escalona¡vel e reproduzavel de ondas gravitacionais.
"Neste estudo, usamos o poder combinado da IA ​​e da supercomputação para ajudar a resolver experimentos de big data relevantes e oportunos. Agora estamos tornando os estudos de IA totalmente reproduzaveis, não apenas averiguando se a IA pode fornecer uma nova solução para grandes desafios, "
Eliu Huerta
Esta nova estrutura indica que os modelos de IA podem ser tão sensaveis quanto os algoritmos de correspondaªncia de modelos tradicionais, mas ordens de magnitude mais rápidos. Além disso, esses algoritmos de IA exigiriam apenas uma unidade de processamento gra¡fico (GPU) de baixo custo, como as encontradas em sistemas de videogame, para processar dados LIGO avana§ados mais rápido do que em tempo real.
O conjunto de IA usado para este estudo processou um maªs inteiro - agosto de 2017 - de dados LIGO avana§ados em menos de sete minutos, distribuindo o conjunto de dados em 64 GPUs NVIDIA V100. O conjunto de IA usado pela equipe para esta análise identificou todas as quatro fusaµes de buracos negros binários previamente identificadas nesse conjunto de dados e não relatou nenhuma classificação incorreta.
"Como cientista da computação, o que éempolgante para mim sobre este projeto", disse Ian Foster, diretor da divisão de Data Science and Learning (DSL) da Argonne, "éque ele mostra como, com as ferramentas certas, os manãtodos de IA podem ser integrados naturalmente em os fluxos de trabalho dos cientistas - permitindo-lhes fazer seu trabalho mais rápido e melhor - aumentando, não substituindo, a inteligaªncia humana. "
Â
Trazendo recursos daspares para suportar, esta equipe interdisciplinar e multi-institucional de colaboradores publicou um artigo na Nature Astronomy apresentando uma abordagem baseada em dados que combina os recursos coletivos de supercomputação da equipe para permitir a detecção de ondas gravitacionais reproduzaveis, aceleradas e baseadas em IA.
"Neste estudo, usamos o poder combinado da IA ​​e da supercomputação para ajudar a resolver experimentos de big data relevantes e oportunos. Agora estamos tornando os estudos de IA totalmente reproduzaveis, não apenas averiguando se a IA pode fornecer uma nova solução para grandes desafios, "Disse Huerta.
Com base na natureza interdisciplinar deste projeto, a equipe espera novas aplicações dessa estrutura baseada em dados, além dos desafios de big data na física.
"Esses investimentos da NSF contem ideias originais e inovadoras que prometem transformar a maneira como os dados cientaficos que chegam em fluxos rápidos são processados. As atividades planejadas estãotrazendo tecnologia de computação acelerada e heterogaªnea para muitas comunidades cientaficas de prática ",
Manish Parashar
"Este trabalho destaca o valor significativo da infraestrutura de dados para a comunidade cientafica ", disse Ben Blaiszik, um cientista pesquisador da Argonne e da Universidade de Chicago. "Os investimentos de longo prazo feitos pelo DOE, a National Science Foundation (NSF), os Institutos Nacionais de Padrões e Tecnologia e outros criaram um conjunto de blocos de construção. a‰ possível para nosreunir esses blocos de construção em maneiras novas e empolgantes de dimensionar essa análise e ajudar a fornecer esses recursos para outras pessoas no futuro. "
Huerta e sua equipe de pesquisa desenvolveram sua nova estrutura com o apoio do NSF, do programa de Pesquisa e Desenvolvimento Dirigido do Laborata³rio de Argonne (LDRD) e do programa de Impacto Computacional Inovador e Novo do DOE em Teoria e Experimento (INCITE).
"Esses investimentos da NSF contem ideias originais e inovadoras que prometem transformar a maneira como os dados cientaficos que chegam em fluxos rápidos são processados. As atividades planejadas estãotrazendo tecnologia de computação acelerada e heterogaªnea para muitas comunidades cientaficas de prática ", disse Manish Parashar, diretor da o Office of Advanced Cyberinfrastructure da NSF.