iMARS: An In-Memory-Computing Architecture for Recommendation Systems

• DOI: 10.48550/arxiv.2202.09433
• Publication Date: 2022-01-01
• Authors: Mengyuan Li; Ann Franchesca La...; Dayane Reis; Xunzhao Yin; Michael Niemier; Xiaobo Sharon Hu
• Keywords: MovieLens; Table (database)

Recommendation systems (RecSys) suggest items to users by predicting their preferences based on historical data. Typical RecSys handle large embedding tables and many table related operations. The memory size bandwidth of the conventional computer architecture restrict performance RecSys. This work proposes an in-memory-computing (IMC) (iMARS) for accelerating filtering ranking stages deep neural network-based iMARS leverages IMC-friendly implemented inside a ferroelectric FET IMC fabric. Circuit-level system-level evaluation show that \fw achieves 16.8x (713x) end-to-end latency (energy) improvement compared GPU counterpart MovieLens dataset.