No abstract

No abstract

This paper presents a ferroelectric FET (FeFET)-based processing-in-memory (PIM) architecture to accelerate the inference of deep neural networks (DNNs). We propose a digital in-memory vector-matrix multiplication (VMM) engine design utilizing the FeFET crossbar to enable bit-parallel computation and eliminate analog-to-digital conversion in prior mixed-signal PIM designs. A dedicated hierarchical network-on-chip (H-NoC) is developed for input broadcasting and on-the-fly partial results processing, reducing the data transmission volume and latency. Simulations in 28-nm CMOS technology show 115× and 6.3× higher computing efficiency (GOPs/W) over desktop GPU (Nvidia GTX 1080Ti) and resistive random access memory (ReRAM)-based design, respectively. INDEX TERMS Deep neural network (DNN), ferroelectric FET (FeFET), processing-in-memory (PIM).

We present an all-optical method for achieving molecular Bose-Einstein condensates of 6 Li. We demonstrate this with mixtures in the lowest two (1-2), and second lowest two (2-3) hyperfine states. For the 1-2 mixture, we can achieve condensate fractions of 36%, with 9 × 10 4 atoms at 0.05 µK temperature. For the 2-3 mixture, we have 28% condensed with 3.2×10 4 atoms at 0.05 µK temperature. We use mostly standard methods, but make a number of refinements in the magnetic bias coils compared with earlier work. Our method imposes minimal constraints on subsequent experiments by allowing plenty of optical access while requiring only one high-vacuum chamber.We use an optical system designed around minimizing the number of active elements, and we can accomplish slowing and sub-Doppler cooling with a single tapered amplifier. * cparker@gatech.edu 1 arXiv:1809.06465v1 [physics.atom-ph]