Compute-In-Memory APU Achieves GPU-Class AI Performance at a Fraction of the Energy Cost

GPU-class performance – The Gemini-I APU delivered comparable throughput to NVIDIA’s A6000 GPU on RAG workloads. Massive energy advantage – The APU delivers over 98% lower energy consumption than a ...
IEEE

Hard Adversarial Example Mining for Improving Robust Fairness

Abstract: Adversarial training (AT) is widely considered the state-of-the-art technique for improving the robustness of deep neural networks (DNNs) against adversarial examples (AEs). Nevertheless, ...