Flexible Computing Systems For AI Acceleration At The Extreme Edge Of The IoT

Embedding intelligence in extreme edge devices allows distilling raw data acquired from sensors into actionable information, directly on IoT end-nodes. This computing paradigm, in which end-nodes no longer depend entirely on the Cloud, offers undeniable benefits, driving a large research area (TinyML) to deploy leading Machine Learning (ML) algorithms on micro-controller class of devices. To fit the limited memory storage capability of these tiny platforms, full-precision Deep Neural Networks (DNNs) are compressed by representing their data down to byte and sub-byte formats, in the integer domain. However, the current generation of micro-controller systems can barely cope with the computing requirements of QNNs. This thesis tackles the challenge from many perspectives, presenting solutions both at software and hardware levels, exploiting parallelism, heterogeneity and software programmability to guarantee high flexibility and high energy-performance proportionality. The first contribution, PULP-NN, is an optimized software computing library for QNN inference on parallel ultra-low-power (PULP) clusters of RISC-V processors, showing one order of magnitude improvements in performance and energy efficiency, compared to current State-of-the-Art (SoA) STM32 micro-controller systems (MCUs) based on ARM Cortex-M cores. The second contribution is XpulpNN, a set of RISC-V domain specific instruction set architecture (ISA) extensions to deal with sub-byte integer arithmetic computation. The solution, including the ISA extensions and the micro-architecture to support them, achieves energy efficiency comparable with dedicated DNN accelerators and surpasses the efficiency of SoA ARM Cortex-M based MCUs, such as the low-end STM32M4 and the high-end STM32H7 devices, by up to three orders of magnitude. To overcome the Von Neumann bottleneck while guaranteeing the highest flexibility, the final contribution integrates an Analog In-Memory Computing accelerator into the PULP cluster, creating a fully programmable heterogeneous fabric that demonstrates end-to-end inference capabilities of SoA MobileNetV2 models, showing two orders of magnitude performance improvements over current SoA analog/digital solutions.

Essays in macroeconomics: insurance, market power, and inequality

This thesis investigates a broad range of topics related to insurance, market power, and inequality, both from an empirical and a theoretical perspective. In the first chapter, I exploit the significant heterogeneity of the shocks hitting Ethiopian households and their heterogeneous response, using relatively recent data (World Bank's LSMS-ISA for households and satellite data for weather shocks). On the one hand, households seem able to insure against most idiosyncratic and mild adverse weather shocks. On the other hand, vulnerability to stronger weather shocks (especially droughts) remains elevated. In the second chapter, starting from firms' individual data, aggregate trends about industry concentration and other proxies of competition are built. This chapter is part of a larger project conducted at the OECD in the Productivity Innovation and Entrepreneurship Division of the STI Directorate The project innovates on the existing literature in its measurement of concentration, aimed at reflecting markets more accurately. On average, aggregate concentration is found to be increasing. In the third chapter, which only lays out some preliminary steps of a more extensive inquiry, I model the heterogeneous effects of aggregate technological progress on individual economic agents and show how this can affect aggregate inequality and other aggregate indicators studied in the macroeconomics literature, such as the entrepreneurship rate and the overall firm distribution. It should be noted, however, that this note is a simple exposition of a possible modelling device rather than a full explanation of these phenomena.