Low-Area and High-Speed Approximate Matrix-Vector Multiplier

摘要: Matrix multiplication is a high-cost operation that can benefit from efficient hardware implementation. Approximate computing offers promising way to lower the costs and speed up computation by leveraging error tolerance of applications like image processing. This work proposes novel approximate matrix-vector multiplier which features low area high speed. Moreover, its accuracy dynamically reconfigurable, allowing user trade small errors for increased Compared previous designs, our approach reduces cost 70% with 5% average error. With more relaxed 10% constraint, it achieves speedup 2x. We apply proposed design color transformation, basic in face-detection algorithms. The transformed images exhibit only decrease detection accuracy.