Approximate Computing exploits the inherent error resilience of machine learning, multimedia, image processing, and wireless communication workloads to trade minimal numerical accuracy for substantial reductions in silicon area, critical-path delay, and switching power.
At I²CS Research Lab, IIIT Kottayam, we design novel approximate 4:2 compressors, Booth and Mitchell logarithmic multipliers, speculative adders, and precision-configurable MAC units—validated both in industry CMOS standard-cell libraries (Synopsys/Cadence) and inside end-to-end deep neural network accelerators.