This paper introduces the Field-Programmable Learning Array, a new paradigm for rapid prototyping of learning primitives and machinelearning algorithms in silicon. The FPLA is a mixed-signal counterpart to the all-digital Field-Programmable Gate Array in that it enables rapid prototyping of algorithms in hardware. Unlike the FPGA, the FPLA is targeted directly for machine learning by providing local, parallel, online analog learning using floating-gate MOS synapse transistors. We present a prototype FPLA chip comprising an array of reconfigurable computational blocks and local interconnect. We demonstrate the viability of this architecture by mapping several learning circuits onto the prototype chip. 1

Abstract—An accurate CMOS current source for current-mode low-voltage differential transmitter drivers has been designed and fabricated. It is composed of binary weighted current mirrors with built-in self-calibration circuits. The proposed self-measurement and calibration circuits can calibrate upon the collective effects of different error contributors due to process, power supply, and temperature variations. The design has been fabricated in standard 0.35- m CMOS technology. Measurement results show that the differential output voltage can be self-calibrated to 1 % accuracy with 16 % reference current variation, 60 % power supply variation, or 13 % load resistance variation, respectively. Index Terms—Built-in self-measurement, CMOS integrated circuits, current-mode transmitter driver, low-voltage, self-calibration. I.