Computer Science > Emerging Technologies
[Submitted on 1 Sep 2026]
Title:FALCON: Fault-Tolerant Magnetic Tunnel Junction-Based In-Memory Stochastic Architecture for Reliability-Critical Edge AI Applications
View PDF HTML (experimental)Abstract:As modern data-centric applications such as neural inference and sensor-edge analytics expand, they increasingly encounter the von Neumann memory wall, suffering from excessive data movement overhead and stringent energy constraints. In-Memory Computing (IMC) utilizing emerging non-volatile technologies, such as Magnetic Tunnel Junctions (MTJs), promises to mitigate these bottlenecks. However, conventional binary radix-based IMC architectures suffer from excessive vulnerability to process-induced variations, restricted operating margins, and thermal noise. To bridge the gap between energy efficiency and computational reliability, this work proposes FALCON, a fault-tolerant, MTJ-based in-memory arithmetic architecture integrated with Stochastic Computing (SC). By encoding numerical values into uniform bit-streams, SC naturally absorbs localized soft errors and enables the execution of an essential suite of arithmetic operations using highly compact logic primitives directly within the memory arrays. FALCON integrates a deterministic bit mapping mechanism with reconfigurable logic-in-memory (LIM) structures, eliminating the need to transfer data to external processors or area- and power-hungry random number generators. Experimental results using 14 nm FinFET technology validate the correct functionality of FALCON even under aggressive voltage scaling, severe process variation, and noise injection levels up to 30%, making it a robust framework for reliability-critical edge AI applications. We investigate the proper functionality of FALCON on morphological closing as a realistic noise-tolerant image processing case study.
Current browse context:
cs.ET
References & Citations
Loading...
Bibliographic and Citation Tools
Bibliographic Explorer (What is the Explorer?)
Connected Papers (What is Connected Papers?)
Litmaps (What is Litmaps?)
scite Smart Citations (What are Smart Citations?)
Code, Data and Media Associated with this Article
alphaXiv (What is alphaXiv?)
CatalyzeX Code Finder for Papers (What is CatalyzeX?)
DagsHub (What is DagsHub?)
Gotit.pub (What is GotitPub?)
Hugging Face (What is Huggingface?)
ScienceCast (What is ScienceCast?)
Demos
Recommenders and Search Tools
Influence Flower (What are Influence Flowers?)
CORE Recommender (What is CORE?)
arXivLabs: experimental projects with community collaborators
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Facts Only
* The work proposes FALCON, a fault-tolerant, MTJ-based in-memory arithmetic architecture integrated with Stochastic Computing (SC).
* IMC utilizing MTJs aims to mitigate data movement overhead and energy constraints in applications like neural inference.
* FALCON encodes numerical values into uniform bit-streams using SC to absorb soft errors.
* FALCON integrates a deterministic bit mapping mechanism with reconfigurable logic-in-memory (LIM) structures.
* Experimental results were validated using 14 nm FinFET technology.
* The architecture functions under aggressive voltage scaling, severe process variation, and noise injection levels up to 30%.
* A case study investigates FALCON functionality on morphological closing for image processing.
Executive Summary
Full Take
The proposal frames reliability as a fundamental constraint when deploying advanced computation at the edge, moving beyond mere efficiency goals. The core innovation lies in shifting error handling from external correction mechanisms to inherent architectural properties through Stochastic Computing—a paradigm shift that uses uncertainty constructively rather than treating it as noise to be eliminated. Integrating fault tolerance directly into the memory array using physical structures (MTJs and LIM) suggests a design philosophy where resilience is built at the hardware level, rather than layered on top of traditional von Neumann models. The focus on mitigating process variation and thermal noise in MTJ systems points toward significant material science challenges regarding device uniformity at small scales. The validation against aggressive operating conditions demonstrates that the proposed solution addresses real-world reliability concerns inherent to next-generation semiconductor manufacturing. The implication is that future edge AI viability depends not just on density and energy efficiency, but on architecting computations where physical stochasticity is leveraged instead of feared as a failure mode.
BRIDGE QUESTIONS: What are the bounds on error tolerance when scaling the complexity of the arithmetic operations within the LIM structures? How can the mathematical properties of morphological closing be systematically mapped onto the probabilistic nature of SC to quantify reliability gains directly? What are the long-term material science implications for MTJ stability under extreme noise injection rates in future edge devices?