Distributed weight update for backpropagation of a neural network

This invention aims to speed up the training of artificial intelligence systems, known as neural networks. It works by having multiple processing units simultaneously update different parts of the AI's internal settings, called weights. After each unit calculates its portion of the updates, these separate contributions are then combined to complete the training step.

Why it matters: Since 2019, the scale of AI models like large language models has grown exponentially, making efficient distributed training a critical bottleneck. This approach to parallelizing weight updates is more essential than ever for training today's massive neural networks.

Status: Not in forceListed as no longer active. The specific reason is not in the record we hold.
How hard to build: SpecializedDistributed computing software, neural network expertise

3 AI reinvention directions for this patent

See all three

AI gives you a few directions you could take this. Pick one, and we check whether your version is different enough to patent, then write the filing.

Reinvent this with AI

Where this one came from

Every morning we go through millions of abandoned patents and pick out 15 worth your time. This was one of today’s.

This one

+10