41530 research outputs found
Sort by
Scalable Hybrid Neuromorphic Accelerator & Hybrid Neural Networks
Get PDFWith machine learning workloads currently at very large scales, models are distributed across large compute systems. On distributed systems, the performance of these models are limited by the bandwidth limitations of chip-to-chip communication. To relieve this bottleneck, spiking neural networks (SNNs) can be utilized to reduce inter-chip communication traffic utilizing inherit network sparsity. However, in comparison to traditional artificial neural networks (ANNs), SNNs can have significant degradation in performance with increased network scale and complexity.
This research proposes a hybrid neural network accelerator that uses the best of both spiking and non-spiking layers by allocating a majority of resources to nonspiking layers on the interior of the chip while bandwidth-limited areas (e.g., I/O pads, or chip separation boundaries) employ spike-based data traffic. By limiting the overall use of spiking layers within the network, we realize the energy savings of SNNs without the a degradation in accuracy which comes with large spike-based models.
We present a scalable chiplet architecture and show how hybrid data is managed with both spike and non-spiking data communication. We also demonstrate how the asynchronous spike-based model is integrated efficiently with the synchronous artificial-based deep learning workloads. We demonstrate that our hybrid architecture offers significant improvements in performance, accuracy, and energy consumption in comparison to SNNs and ANNs. With up to a 1.34× increase in energy efficiency and 1.56× decrease in single inference latency, the versatility of the architecture is demonstrated by its validation across multiple datasets, encompassing both language processing and computer vision tasks
Automatic Mass Balancing Of A Spacecraft Attitude Dynamics Simulator With Six Sliding Masses
Get PDFThe goal of this thesis is to investigate automatic mass balancing methods for spacecraft attitude dynamics simulators, create a hardware design for a mass balancing system, and assemble the hardware on the Cal Poly Spacecraft Attitude Dynamics Simulator (SADS). Spacecraft attitude dynamics simulators replicate the torque-free environment of space with ground-based hardware. The SADS is mounted on a spherical air bearing, and includes a pyramid of four reaction wheels. The air bearing allows frictionless, unbounded rotation about the vertical axis, and 30 degrees about the horizontal axes. The torque-free configuration of the SADS can be used to test spacecraft attitude control software and hardware. For spacecraft attitude simulators, it is essential to accurately align the center of mass with the center of rotation. Any alignment error will cause a gravitational disturbance torque that quickly saturates actuators. It is necessary to use an automatic mass balancing system if the system\u27s hardware is regularly adjusted, since manual mass balancing methods become prohibitively time consuming. Once balanced, it is also useful if the attitude simulator can measure its inertia tensor for use in control software. To avoid displacing the center of mass, a symmetrical six-sliding-mass balancing system was developed for the SADS, driven by geared DC motors. Several mass balancing algorithms were designed that do not require additional actuators. These algorithms enable mass balancing prior to the integration of reaction wheels, and are intended to eventually be part of a two-stage balancing approach. In the controlled stage, a continuous time controller corrects the gravitational disturbance torque. A moment distribution law for the sliding masses is then used with an FSFB or adaptive controller to align the center of mass and rotation. In the system identification stage, angles and angular rates are measured during phased torque sine sweeps so that the inertia tensor and center of mass position can be estimated. After making a correction to the sliding mass positions, the identification and correction process is repeated until the system is balanced. To be considered \u27balanced,\u27 any remaining disturbance torques must be negligible on the time scale of a control system simulation. The horizontal axis positional errors for the center of mass should be on the order of one tenth of a millimeter, and it is generally acceptable for the vertical axis error to be approximately an order of magnitude greater. Since a vertical center of mass offset is generally aligned with the structure of the air bearing, its corresponding gravity disturbance torque is reduced. Since the mass properties of the SADS must be measured regardless, they are a convenient metric for the effectiveness of the balancing system
A Study On Privacy Over Security And Privacy Enhancing Networks
Get PDFWith rapid developments in communication technologies and awareness of security and privacy risks online, Security and Privacy Enhancing Networks (SPENs) have become increasingly popular. Especially during the COVID-19 pandemic, workplaces encouraged employees to take additional security measures, such as VPNs. In this work, we conduct a comprehensive study on website fingerprinting attacks. A comprehensive system model and threat model based on two types of SPENs (Virtual Private Networks and Tor Networks) are presented. Moreover, we demonstrate a website fingerprinting attack by ethically collecting website fetch data and analyzing the collected data using five different machine learning classification models including k nearest neighbors, decision tree, ada boost, and random forest. We find that SPENs are still vulnerable to website fingerprinting attacks which enable attackers to violate users’ behavioral privacy. However, it is not easy to get accurate results, especially over a large number of websites. Furthermore, we discuss a series of recommendations for SPENs to increase behavioral privacy for their customers. Finally, we cover a variety of directions that future work could take
Thinking Styles in the Built Environment
Get PDFEffective communication in the built environment is crucial for project success, information relay, goal alignment, error prevention, and safety. Diverse parties, including general contractors, architects, engineers, and subcontractors, bring varying Thinking Styles to the table, as defined by Harrison and Bramson. These styles—Synthesist, Idealist, Pragmatist, Analyst, and Realist—impact collaboration and conflict within projects. This study aims to explore where thinking differs among these parties and the congruity of Thinking Styles within each group. Results of this study indicate a predominant Realist Thinking Style among general contractors and subcontractors with subcontractors equally leaning towards Analyst. Engineers, although expected to be Analysts, primarily identified as Pragmatists. Architects displayed a diverse range of thinking styles, uniquely including the only Synthesist in the study, reflecting their creative and often controversial approach. The findings reveal notable correlations between Thinking Styles and the Built Environment’s roles, suggesting that knowledge of these styles can enhance teamwork and mitigate conflicts. Realists and Analysts were seen as the easiest to work with, while Synthesists and Idealists were more challenging, highlighting the need for tailored communication strategies in diverse teams. This study emphasizes the importance of understanding Thinking Styles to improve collaboration and efficiency in the built environment
El nacimiento del rock en español: ¿Cómo fue que este género de música latina tuvo éxito?
No full textThis paper titled El nacimiento del rock en español: ¿Cómo fue que este género de música latina tuvo éxito? will cover information regarding the Spanish rock movement. This essay compares what is written in this paper to that of prior scholars to show that although they do bring valid points, there is a lot more than many of us realize. There are various articles and reports written about the movement, but many do not explain the creation of the movement, let alone its survival in the music industry as a new genre that some had never heard of or had no faith in. This research paper will focus on how this came to be in Mexico where many famous bands originated from, as well as how the United States perceived this new music in the market. I aim to show the differing perspectives people had, as well as explain the genre’s huge marketing success.
I will also analyze the bands themselves, explaining their distinct styles and why they became popular. I will show how they promoted this musical movement inside and outside of the United States through my analysis of interviews as well as articles from different periods explaining the impacts they had inside and outside of the Latino/x communities. In the end, I will add my own experiences to confirm some of the details mentioned in these articles and interviews, as well as give my firsthand account of what it looks like to be at one of these concerts
NSF-ERI Evolution of Liquid Sheet Edges Project: Soap Film Frame Withdrawal and Positioning Apparatus
Get PDFThe experimental investigation of instabilities along retracting liquid sheet edges seeks to answer a fundamental fluid mechanics question. We’ve successfully delivered an experimental apparatus to conduct research and answer this question. This document chronicles a design overview of our apparatus, our manufacturing process, our design verification, and recommendations for future steps. The final product is a document outlining our final design (i.e., our ‘Final Design Review’ (FDR))
Advanced Wildfire Hose Clamp
Get PDFThis report describes in detail the manufacturing, assembly, and testing of our hose clamp. We start with a review of the changes we have made to our design since our last report, namely a new automatic latching system, and continue with our manufacturing and assembly processes. Then we detail our testing procedures that we created to evaluate how well our design meets user needs and discuss the results of our tests. Unfortunately, our design did not meet our residual flow rate criteria and showed excessive leaking when clamped, so we end with some discussion of the issues our design has, and recommendations for how they may be remedied when this project is continued next year