channel quantization
Recently Published Documents


TOTAL DOCUMENTS

82
(FIVE YEARS 13)

H-INDEX

12
(FIVE YEARS 1)

Electronics ◽  
2021 ◽  
Vol 10 (22) ◽  
pp. 2823
Author(s):  
Maarten Vandersteegen ◽  
Kristof Van Beeck ◽  
Toon Goedemé

Quantization of neural networks has been one of the most popular techniques to compress models for embedded (IoT) hardware platforms with highly constrained latency, storage, memory-bandwidth, and energy specifications. Limiting the number of bits per weight and activation has been the main focus in the literature. To avoid major degradation of accuracy, common quantization methods introduce additional scale factors to adapt the quantized values to the diverse data ranges, present in full-precision (floating-point) neural networks. These scales are usually kept in high precision, requiring the target compute engine to support a few high-precision multiplications, which is not desirable due to the larger hardware cost. Little effort has yet been invested in trying to avoid high-precision multipliers altogether, especially in combination with 4 bit weights. This work proposes a new quantization scheme, based on power-of-two quantization scales, that works on-par compared to uniform per-channel quantization with full-precision 32 bit quantization scales when using only 4 bit weights. This is done through the addition of a low-precision lookup-table that translates stored 4 bit weights into nonuniformly distributed 8 bit weights for internal computation. All our quantized ImageNet CNNs achieved or even exceeded the Top-1 accuracy of their full-precision counterparts, with ResNet18 exceeding its full-precision model by 0.35%. Our MobileNetV2 model achieved state-of-the-art performance with only a slight drop in accuracy of 0.51%.


2021 ◽  
Author(s):  
Daniel L. Colon ◽  
Mariano E. Chicatun ◽  
Fernando H. Gregorio ◽  
Juan E. Cousseau

2021 ◽  
Vol 5 (3) ◽  
pp. 1-22
Author(s):  
Kai Li ◽  
Ning Lu ◽  
Jingjing Zheng ◽  
Pei Zhang ◽  
Wei Ni ◽  
...  

Thanks to flexible deployment and excellent maneuverability, autonomous drones have been recently considered as an effective means to act as aerial data relays for wireless ground devices with limited or no cellular infrastructure, e.g., smart farming in a remote area. Due to the broadcast nature of wireless channels, data communications between the drones and the ground devices are vulnerable to eavesdropping attacks. This article develops BloothAir, which is a secure multi-hop aerial relay system based on Bluetooth Low Energy ( BLE ) connected autonomous drones. For encrypting the BLE communications in BloothAir, a channel-based secret key generation is proposed, where received signal strength at the drones and the ground devices is quantized to generate the secret keys. Moreover, a dynamic programming-based channel quantization scheme is studied to minimize the secret key bit mismatch rate of the drones and the ground devices by recursively adjusting the quantization intervals. To validate the design of BloothAir, we build a multi-hop aerial relay testbed by using the MX400 drone platform and the Gust radio transceiver, which is a new lightweight onboard BLE communicator specially developed for the drone. Extensive real-world experiments demonstrate that the BloothAir system achieves a significantly lower secret key bit mismatch rate than the key generation benchmarks, which use the static quantization intervals. In addition, the high randomness of the generated secret keys is verified by the standard NIST test, thereby effectively protecting the BLE communications in BloothAir from the eavesdropping attacks.


Complexity ◽  
2020 ◽  
Vol 2020 ◽  
pp. 1-15 ◽  
Author(s):  
Tian Tan ◽  
Jinfeng Gao ◽  
Jinxia Wang ◽  
Zhen Zhao ◽  
Miao Ma

This paper investigates the hybrid-driven mechanism problem for Markov jump system, where both channel quantization (BCQ) and network-induced delay based on uncertain network are considered. Firstly, comparing with the traditional event-triggered scheme, a hybrid-driven mechanism is employed in networked control systems (NCSs) for the finite capacity of communication bandwidth resources and system performance in equilibrium. Then, the quantization technology is applied in the communication channel from sensor-to-controller and controller-to-actuator. The application of BCQ is for further investigation that mitigate data packet transmission rate. Thirdly, Markov jump system is modeled for the hybrid-driven mechanism and network-induced delay. By constructing the Lyapunov–Krasovskii function, a sufficient condition is derived as the stability criterion, and the controller is designed in which the nonlinear term is rewritten for simplifying the calculation. Finally, two simulation examples are provided to demonstrate the effectiveness of the proposed approach.


2020 ◽  
pp. 791-835
Author(s):  
Giuseppe Mussardo

The Thermodynamic Bethe Ansatz (TBA) allows us to study finite size and finite temperature effects of an integrable model. This chapter investigates the integral equations that determine the free energy and gives their physical interpretation. It discusses Casimir energy, Bethe relativistic wave function, the derivation of thermodynamics, the meaning of pseudo-energy (dressed energy and momentum), infrared and ultraviolet limits, the coefficient of bulk energy, the general form of the TBA equations, the thermodynamics of the free field theories, L-channel quantization and the LeClair–Mussardo formula. It also covers the application of the Yang–Lee S-matrix, the magnetic field Ising model, and the tricritical Ising model.


Sign in / Sign up

Export Citation Format

Share Document