Search papers, labs, and topics across Lattice.
This paper presents RACE-AIMC, a framework that optimally selects a single analog in-memory computing (AIMC) accelerator from a pool of heterogeneous devices while providing a certified upper bound on its error rate. By leveraging statistical methods, RACE-AIMC enables energy-efficient inference by activating only the best-performing chip for a given energy budget and deferring to a fallback only when necessary. The results show that RACE-AIMC achieves a mean error bound of 7.83% while reducing energy consumption by 69.02% compared to running all accelerators simultaneously, matching the accuracy of traditional digital systems.
RACE-AIMC cuts energy use by 69% while maintaining accuracy comparable to digital systems by intelligently selecting the best AIMC accelerator and certifying its reliability.
Analog in-memory computing (AIMC) speeds up neural-network inference by doing the arithmetic directly inside a memory array, instead of shuttling weights back and forth between memory and a processor. This saves energy, but the physical devices that store the weights are imperfect: programming errors, electrical noise, limited-resolution converters, and outright broken cells all distort the computation, and every physical chip is distorted in its own way. A designer with several such chips available faces an uncomfortable choice: run all of them and combine the answers (safe, but wasteful of energy), or trust a single chip blindly (cheap, but with no guarantee on how often it is wrong). This paper introduces RACE-AIMC (Risk-Aware Certified Ensemble for AIMC), a framework that resolves this choice with statistics rather than guesswork. Offline, RACE-AIMC studies a pool of physical accelerators, picks the single best one for a given energy budget, and computes a mathematically exact upper bound on how often that accelerator will be wrong when it chooses to answer. Online, only that one accelerator is switched on; a lightweight check decides whether to accept its answer or defer to a fallback. In our simulations using a noisy weight mapping and multiple independent test runs, every certified bound stayed under a 10% error target (mean bound 7.83% +- 0.89%, with 70.88% +- 0.98% of inputs answered directly). The resulting system matches the accuracy of a clean digital baseline while cutting modeled energy use by 69.02% relative to always running every accelerator in the pool.
