Training of convolutional neural networks (CNNs) on embedded platforms to support on-device learning is earning vital importance in recent days. Designing flexible training hardware is much more challenging than inference hardware, due to design complexity and large computation/memory requirement. In this work, we present an automatic compiler based FPGA accelerator with 16-bit fixed-point precision for complete CNN training, including Forward Pass (FP), Backward Pass (BP) and Weight Update (WU). We implemented an optimized RTL library to perform training-specific tasks, and developed an RTL compiler to automatically generate FPGA-synthesizable RTL based on user-defined constraints. We present a new cyclic weight storage/access scheme for on-chip BRAM and off-chip DRAM to efficiently implement non-transpose and transpose operations during FP and BP phases, respectively. Representative CNNs for CIFAR-10 dataset are implemented and trained on Intel Stratix 10 GX FPGA using proposed hardware architecture, demonstrating up to 479 GOPS performance.
Multimedia processing is becoming increasingly important with a wide variety of applications ranging from multimedia cell phones to high-definition interactive television. Media processing involves the capture, storage, manipulation and transmission of multimedia objects such as text, handwritten data, audio objects, still images, 2-D/3-D graphics, animation, and full-motion video. A number of implementation strategies have been proposed for processing multimedia data. These approaches can be broadly classified based on the evolution of processing architectures and the functionality of the processors. In order to provide media processing solutions to different consumer markets, designers have combined some of the classical features from both the functional and evolution-based classifications resulting in many hybrid solutions. We propose a categorization of existing microprocessors based on a combination of both architectural and functional flavors with examples of each approach from the latest multimedia processing families. The varying processing requirements in multimedia computing for reconfigurable multimedia processing.
Executive summaryThis report presents the results of an analysis of published studies and other literature concerned with terminal box control, occupancy sensing technology, and multi-zone demand control ventilation (DCV) in commercial buildings. To meet the ventilation needs of building occupants, heating, ventilating and airconditioning systems provide outdoor air. For many systems, air is brought into a commercial building through air-handling units, which supply conditioned air to many thermal zones in the building. The airhandling units mix outdoor air in a controlled proportion with recirculated air and then cool the mixture before distribution to the terminal boxes. Terminal boxes usually serve a single building zone, controlling the air flow rate to the zone and reheating the air, if it is too cool for the zone served. Each terminal box has a minimum air flow rate set so that it meets the ventilation requirements of the occupants of the zone the box serves. This minimum air flow rate is commonly designed as a constant value based on the design occupancy of the zone served, which usually corresponds to the maximum occupancy.In practice, control system integrators and installers often set the minimum air flow rate for ventilation to between 30% and 50% of the terminal box maximum air flow rate. Building occupancy, however, varies dynamically. Conference rooms, cafeterias, break rooms, auditoriums, and other assembly spaces are often unoccupied for significant periods of time. Office occupancy varies during the course of a work day, from day to day, and over the longer term because of meetings in the office, attendance of meetings elsewhere, business travel, changing room functions during remodeling, and variations in staffing. The resulting over-ventilation, during times when the space has less than maximum occupancy or is unoccupied, wastes significant fan power, resulting in energy waste, and even causing discomfort for occupants in some spaces (e.g., conference rooms) from overcooling.Carbon-dioxide-based demand control ventilation has existed for a number of years; however, as currently used, it controls outdoor-air intake by the air handler only, not ventilation of individual zones or rooms. The need for frequent recalibration (e.g., annually) and the high cost of installation for each zone (or room) makes CO 2 -based control of ventilation at terminal boxes impractical. A multi-lab report (Brambley et al. 2005) for the U.S. Department of Energy (DOE) on advanced sensors and controls identifies a need for "development of sensors to determine occupancy number and population distribution within buildings." There are no such terminal unit controllers on the market today to modulate air flow to zones based on actual occupancy. The Advanced Energy Retrofit Guide (AERG) for Office Buildings, sponsored by DOE's Building Technology Program (BTP), reveals that a 4% reduction in whole-building energy consumption can be achieved by implementing occupancy-based control (OBC) in conference rooms alone ).This repor...
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.