Reproduce “In-Memory Nearest Neighbor Search with FeFET Multi-Bit Content-Addressable Memories” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Arman Kazemi, Mohammad Mehdi Sharifi, Ann Franchesca Laguna, Franz Müller, Ramin Rajaei, Ricardo Olivo, Thomas Kämpfe, Michael Niemier and 1 more
Researchers from the University of Notre Dame and Fraunhofer IPMS-CNT developed FeFET Multi-Bit Content-Addressable Memories (MCAMs) that perform in-memory Nearest Neighbor (NN) search using a novel, native distance function. This approach achieves classification accuracies comparable to floating-point software, outperforming prior hardware accelerators by up to 13%, while demonstrating robustness to device variations and experimental feasibility.
Key insights
- FeFETs' multi-level storage capabilities and exponential transfer characteristics enable the native implementation of a highly discriminative distance function directly within the memory, suitable for accurate NN search.
- MCAMs offer a robust and practical alternative to fully analog CAMs by operating on discrete multi-bit states, bypassing complex analog programming and on-the-fly signal processing issues.
- The proposed FeFET MCAM distance function demonstrates remarkable resilience to device-to-device threshold voltage (Vth) variations, maintaining accuracy even with significant Vth sigma values up to 80mV.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.