Sustainability of Edge AI at Scale: An empirical study on the sustainability of Edge AI in terms of energy consumption

Edge AI is an architectural deployment tactic that brings AI models closer to the user and data, relieving internet bandwidth usage and providing low latency and privacy. It remains unclear how this tactic performs at scale, since the distribution overhead could impact the total energy consumption. We identify four architectural scalability...

Modular Neural Network Navigation for Autonomous Nano Drone Racing

In this study, we present a first step towards a cutting-edge software framework that will enable autonomous racing capabilities for nano drones. Through the integration of neural networks tailored for real-time operation on resource-constrained devices. A lightweight Convolutional Neural Network, with the Gatenet architecture, is adjusted for...

A New Logarithmic Quantization Technique and Corresponding Processing Element Design for CNN Accelerators

Convolutional Neural Networks (CNN) have become a popular solution for computer vision problems. However, due to the high data volumes and intensive computation involved in CNNs, deploying CNNs on low-power hardware systems is still challenging.<br/>The power consumption of CNNs can be prohibitive in the most common implementation platforms:...

Hybrid Posit and Fixed Point Hardware for Quantized DNN Inference

The recently introduced posit number system was designed as a replacement for IEEE 754 floating point, to alleviate some of its shortcomings. As the number distribution of posits is similar to the data distributions in deep neural networks (DNNs), posits offer a good alternative to fixed point numbers in DNNs: using posits can result in high...

Topology of a Cooper Pair Transistor

Superconducting quantum circuits came out as promising candidates for the exploration of topological phenomena that are currently inaccessible in condensed<br/>matter systems. One such circuit is a Cooper pair transistor which has already<br/>been widely studied in different regimes of operation due to its importance in<br/>quantum computation....

Proposed solutions for quantization induced performance deterioration in reset controllers

The most widely applied feedback controller is PID. This controller gains its popularity because of the ease of design through loop shaping, since PID can be analyzed in the frequency domain. However, PID is limited by linearity. Reset control is a nonlinear addition to PID control. Through linearization techniques it can be designed by...

Cryogenic Hardware Considerations Of Neural Network Decoders For Quantum Error Correction Using Rotated Surface Codes

The quantum bits (qubits) at the core of any quantum computers are so fragile that quantum error correction(QEC) schemes are needed to increase their robustness and enable fault-tolerant quantum algorithms. The surface code is one of the most popular QEC schemes, but it requires the availability of an efficient decoder. While neural networks...

Binary Neural Networks for Object Detection

In the past few years, convolutional neural networks (CNNs) have been widely utilized and shown state-of-the-art performances on computer vision tasks. However, CNN based approaches usually require a large amount of storage, run-time memory, as well as computation power in both training and inference time, which are usually used on GPU based...

Stochastic Resonance Analog-to-Digital Conversion: 1-Bit Signal Acquisition Employing Noise

Stochastic resonance (SR) is a phenomenon in which the presence of noise increases the performance of the system. The phenomenon has first been discovered in a climate change model and is later observed in neuronal systems. In artificial, electronic, systems, stochastic resonance is observed in systems based on Schmitt-Triggers and comparators....

Quantized Distributed Optimization Schemes: A monotone operator approach

Recently, the effects of quantization on the Primal-Dual Method of Multipliers were studied.<br/>In this thesis, we have used this method as an example to further investigate the effects of quantization on distributed optimization schemes in a much broader sense. Using monotone operator theory, the effect of quantization on all distributed...

Adaptive quantized control for uncertain networked systems

Major advancements over the last few decades in communication networks gave rise to the<br/>new paradigm of Networked Control Systems (NCSs). Within this paradigm, sensing and<br/>actuation signals are exchanged among various parts of a single system or among many<br/>subsystems via communication networks. Although this enables one to perform...

An Element-Level Sigma-Delta ADC for Ultrasound Imaging

This work presents an architecture capable of digitizing every channel of an ultrasound transducer array independently and simultaneously. This feature is achieved by exploiting the frequency response of the piezoelectric transducer in order to save area and reduce to a minimum the required building blocks of the ADC. The transducer is used as...

Quantization effects in PDMM: A first study for synchronous distributed averaging

Nowadays, large-scale networks of computing units, often characterized by the absence of central control, have become more commonplace in many applications. To facilitate data processing in these large-scale networks distributed signal processing is required. The iterative behaviour of distributed processing algorithms combined with the limited...

Sensor Selection and Bit Allocation in WSNs with Realistic Digital Communication Channels

For energy management in wireless sensor networks, only the sensors with most informative measurements are activated to operate. How to select sensors that make good tradeoff between performance and energy consumption is what many researchers are focusing on. Existing solutions assume analog data model, i.e., the data from sensors collected by a...

An asymmetric speech coding algorithm for non real-time speech applications in multi-media systems

After looking at properties of speech and existing speech coding algorithms, a proposal for a new asymmetric speech coding algorithm is done. An asymmetric algorithm is a system in which the decoder is significantly simpler than the encoder. The purpose of this asymmetry is to exchange encoding delay for a decrease in bitrate. These algorithms...