Fault-Aware Training to Salvage AI Accelerators
Summary
The USPTO published patent application US20260093976A1 by inventor Ihab Amer describing a method for fault-aware training of AI accelerators. The method generates multiple neural network model approximations of a compute engine with faults, matching a fault map to an appropriate approximation, and loading the matched model when the engine transitions to approximate mode. The compute engine is a multiply-accumulate unit within an AI accelerator.
What changed
USPTO published patent application US20260093976A1 for a fault-aware training method to salvage AI accelerators with defective compute engines. The invention generates multiple neural network model approximations of an integrated circuit compute engine containing at least one fault, matches a fault map loaded to the IC with one of these approximations, and loads the matched approximation when the engine transitions to approximate mode. In approximate mode, higher precision arithmetic operations are substituted with lower precision operations. The compute engine is specifically a multiply-accumulate unit within an AI accelerator.
This patent application does not impose compliance obligations on any party. It represents intellectual property protection for a technical method. Entities involved in AI accelerator development or manufacturing may wish to review the claims to assess potential overlap with their own technologies or to consider licensing implications. No deadlines or penalties are associated with this publication.
Archived snapshot
Apr 2, 2026GovPing captured this document from the original source. If the source has since changed or been removed, this is the text as it existed at that time.
FAULT-AWARE TRAINING TO SALVAGE AI ACCELERATORS
Application US20260093976A1 Kind: A1 Apr 02, 2026
Inventors
Ihab AMER
Abstract
Embodiments herein describe a method for generating multiple neural network model approximations of a compute engine of an integrated circuit (IC) including at least one fault, matching a fault map loaded to the IC with one of the multiple neural network model approximations, and loading a matched neural network model approximation to the IC. The compute engine is a multiply-accumulate (MAC) unit incorporated within an artificial intelligence (AI) accelerator. The multiple neural network model approximations are generated when the compute engine transitions into an approximate mode. In the approximate mode, a first set of operations are substituted for a second set of operations, where the first set of operations are higher precision arithmetic operations and the second set of operations are lower precision arithmetic operations.
CPC Classifications
G06N 3/08 G06F 30/327 G06N 3/0475
Filing Date
2024-09-27
Application No.
18900435
Related changes
Get daily alerts for USPTO Patent Applications - AI & Computing (G06N)
Daily digest delivered to your inbox.
Free. Unsubscribe anytime.
Source
About this page
Every important government, regulator, and court update from around the world. One place. Real-time. Free. Our mission
Source document text, dates, docket IDs, and authority are extracted directly from USPTO.
The summary, classification, recommended actions, deadlines, and penalty information are AI-generated from the original text and may contain errors. Always verify against the source document.
Classification
Who this affects
Taxonomy
Browse Categories
Get alerts for this source
We'll email you when USPTO Patent Applications - AI & Computing (G06N) publishes new changes.
Subscribed!
Optional. Filters your digest to exactly the updates that matter to you.