Deep neural networks use skip connections to improve training convergence. However, these skip connections are costly in hardware, requiring extra buffers and increasing on- and off-chip memory utilization and bandwidth requirements. In this article, we show that skip connections can be optimized for hardware when tackled with a hardware-software codesign approach. We argue that while a network’s skip connections are needed for the network to learn, they can later be removed or shortened to provide a more hardware-efficient implementation with minimal to no accuracy loss. We introduce Tailor , a codesign tool whose hardware-aware training algorithm gradually removes or shortens a fully trained network’s skip connections to lower the hardware cost. Tailor improves resource utilization by up to 34% for block random access memories (BRAMs), 13% for flip-flops (FFs), and 16% for look-up tables (LUTs) for on-chip, dataflow-style architectures. Tailor increases performance by 30% and reduces memory bandwidth by 45% for a two-dimensional processing element array architecture.

1.5 billion smartphones are sold annually, and most are decommissioned less than two years later. Most of these unwanted smartphones are neither discarded nor recycled but languish in junk drawers and storage units. This computational stockpile represents a substantial wasted potential: modern smartphones have increasingly high-performance and energy-efficient processors, extensive networking capabilities, and a reliable built-in power supply. This project studies the ability to reuse smartphones as "junkyard computers." Junkyard computers grow global computing capacity by extending device lifetimes, which supplants the manufacture of new devices. We show that the capabilities of even decade-old smartphones are within those demanded by modern cloud microservices and discuss how to combine phones to perform increasingly complex tasks. We describe how current operation-focused metrics do not capture the actual carbon costs of compute. We propose Computational Carbon Intensity---a performance metric that balances the continued service of older devices with the superlinear runtime improvements of newer machines. We use this metric to redefine device service lifetime in terms of carbon efficiency. We develop a cloudlet of reused Pixel 3A phones. We analyze the carbon benefits of deploying large, end-to-end microservice-based applications on these smartphones. Finally, we describe system architectures and associated challenges to scale to cloudlets with hundreds and thousands of smartphones.

Deep neural networks employ skip connections – identity functions that combine the outputs of different layers-to improve training convergence; however, these skip connections are costly to implement in hardware. In particular, for inference accelerators on resource-limited platforms, they require extra buffers, increasing not only on- and off-chip memory utilization but also memory bandwidth requirements. Thus, a network that has skip connections costs more to deploy in hardware than one that has none. We argue that, for certain classification tasks, a network's skip connections are needed for the network to learn but not necessary for inference after convergence. We thus explore removing skip connections from a fully-trained network to mitigate their hardware cost. From this investigation, we introduce a fine-tuning/retraining method that adapts a network's skip connections – by either removing or shortening them-to make them fit better in hardware with minimal to no loss in accuracy. With these changes, we decrease resource utilization by up to 34% for BRAMs, 7% for FFs, and 12% LUTs when implemented on an FPGA.

This paper discusses experiments on soil-based microbial fuel cells (MFCs) as energy scavenging sources. We explain the mechanism of operation for MFCs, perform controlled laboratory experiments of MFCs, and deploy a small-scale insitu pilot in an active farm. We find that traditional energy harvester ICs draw power too aggressively, which reduces overall energy capture. We show that isolated MFCs can be combined in series or parallel to improve the voltage or current output of the harvesting source. Lastly, we observe that under a real-world, drip-irrigated agricultural setting, MFC output is appreciably lower, but consistent at 0.5-2 microwatts.

Soil microbial fuel cells are a promising source of energy for outdoor sensor networks. These biological systems are sensitive to environmental conditions, therefore more data is needed on their behavior "in the wild" to enable the creation of an energy system capable of being widely deployed. Prior work on early characterization of microbial fuel cells relied on extremely accurate, but expensive, logging hardware. To scale up the number of deployment sites, we present custom logging hardware, specially designed to accurately monitor the behavior of microbial fuel cells at low cost. This paper describes the design and evaluation of the board, which is open source and freely available on GitHub.

This demo showcases the power delivery potential of soil-based microbial fuel cells. We build a prototype energy harvesting setup for a soil microbial fuel cell, measure the amount of power that we can harvest, and use that energy to drive an e-ink display. Microbial fuel cells are highly sensitive to environmental conditions, especially soil moisture. In near-optimal, super moist conditions our cell provides approximately 100 μW of power at around 500 mV, which is ample power over time to power our system several times a day. In sum, we find that the confluence of ever lower-power electronics and new understanding of microbial fuel cell design means that "soil-powered sensors" are now feasible. There remains, however, significant future work to make these systems reliable and maximally performant.