CORTEXA
← Browse

Anshul Arunachalam

1 paper indexed

openalexeScholarship (California Digital Library)

Optimizing Ring AllReduce for Sparse Data

Anshul Arunachalam

The distributed training of machine learning models via gradient descent is generally conducted by iteratively computing the local gradients of a loss function and aggregating them across all processors. Communicating these gradients during aggregation is often a major cost but s…

View free PDFSource page