Carry-through checksum is proposed, a fundamentally new scheme for soft-error detection in CNN inference on embedded GPUs that embeds dedicated carry-through filters into the convolutional layers, which compute a checksum from the CNN's own operations and propagate it through inference, enabling end-to-end error detection with a single output verification.
Abstract
Convolutional Neural Networks (CNNs) are increasingly deployed in safety-critical edge applications, where soft errors can silently corrupt inference outputs and lead to unsafe decisions. Such applications typically rely on resource-constrained embedded GPUs, requiring fault detection and mitigation techniques that add minimal compute, memory, and latency overhead while integrating seamlessly with the standard GPU inference pipeline. Existing algorithm-based fault tolerance techniques rely on matrix augmentation and per-operation checksum verification, imposing substantial overhead that is prohibitive for CNN inference on embedded GPUs. In this work, we propose carry-through checksum, a fundamentally new scheme for soft-error detection in CNN inference on embedded GPUs. The method embeds dedicated carry-through filters into the convolutional layers, which compute a checksum from the CNN's own operations and propagate it through inference, enabling end-to-end error detection with a single output verification. Experimental results on multiple CNN architectures show that the proposed method detects 95.86% and 86.56% of critical faults for FP32 and FP16, respectively, at almost no additional per-image overhead. Detected faults are mitigated through re-execution, incurring only 2.27% run-time overhead across the entire test set on an NVIDIA Jetson Orin NX GPU.
The deployment of Convolutional Neural Networks (CNNs) for mechanical fault diagnosis at the edge is often limited by the parameter explosion in fully connected layers when processing high-dimensional spectral features, such as the Fast Fourier Transform (FFT). This paper investigates the feasibility of low-latency inf...
L. Carvalho, Luiz Guilherme Ito, J. R. Azambuja· Symposium on Integrated Circ...· 0 citations
Computer-vision systems for construction safety sit at a design tension: the target device is an embedded GPU with tight latency and power budgets, yet the object space that makes a site truly readable spans personal protective equipment, heavy machinery, and handheld tools. A monolithic detector either shrinks to fit...
Luiz Guilherme Ito, L. Carvalho, J. R. Azambuja· Symposium on Integrated Circ...· 0 citations
Low-precision inference is widely adopted in edge AI to reduce computational cost and memory footprint. However, existing open-source accelerator platforms provide limited end-to-end support for CNNs following the standard TensorFlow Lite integer inference scheme. This paper presents ZTA-Q, an open-source RISC-V-based...
Yi-Ke Li, M. AjayKumar, PS Vishnu et al.· 0 citations
A compact convolutional neural network architecture integrated with a multi-head attention mechanism to enhance feature discrimination in malware family classification that attains high weighted Precision, Recall, and F1 scores and enhanced macro-averaged performance, particularly enhancing classification for minority...
Mohammed Kasem Al-Bayati· Zanco Journal of Pure and Ap...· 0 citations
CNN inference on edge hardware is constrained by memory bandwidth, redundant logic, and the area overhead of fully parallel multiply-accumulate (MAC) units. This paper presents a hardware-efficient CNN accelerator with an SRAM-based architecture optimised for classifying 28×28 grayscale images. Input images are ingeste...
Avinash Krishna Pk, R. S· International Conference Inn...· 0 citations
Related blog posts
Microsoft Research Blog· microsoft.comNov 16, 2026
AI coding agents are increasingly help developers fix bugs and build features largely on their own, but every interaction with their models consumes tokens. In each session, an agent sends instructions, tool descriptions, and session history to its model. Most efforts to reduce this cost focus on individual tasks. Yet developers return to the same […] The post TokenEconomizer: Learning from Experience to Make AI Coding Agents More Efficient appeared first on Microsoft Research.
MIT News · Artificial Intelligence· news.mit.eduNov 13, 2026