Activation Redistribution Based Hybrid Asymmetric Quantization Method of Neural Networks

The demand for adopting neural networks in resource-constrained embedded devices is continuously increasing.Quantization is one of the most promising solutions to reduce computational cost and memory storage on embedded devices.In order to reduce the complexity and overhead of deploying neural networks on Integeronly hardware,most current quantization methods use a symmetric quantization mapping strategy to quantize a floating-point neural network into an integer network.However,although symmetric quantization has the advantage of easier implementation,it is sub-optimal for cases where the range could be skewed and not symmetric.This often comes at the cost of lower accuracy.This paper proposed an activation redistribution-based hybrid asymmetric quantizationmethod for neural networks.The proposedmethod takes data distribution into consideration and can resolve the contradiction between the quantization accuracy and the ease of implementation,balance the trade-off between clipping range and quantization resolution,and thus improve the accuracy of the quantized neural network.The experimental results indicate that the accuracy of the proposed method is 2.02%and 5.52%higher than the traditional symmetric quantization method for classification and detection tasks,respectively.The proposed method paves the way for computationally intensive neural network models to be deployed on devices with limited computing resources.Codes will be available on https://github.com/ycjcy/Hybrid-Asymmetric-Quantization.

Numerical Study on Reduction in Aerodynamic Drag and Noise of High-Speed Pantograph

Reducing the aerodynamic drag and noise levels of high-speed pantographs is important for promoting environmentally friendly,energy efficient and rapid advances in train technology.Using computational fluid dynamics theory and the K-FWH acoustic equation,a numerical simulation is conducted to investigate the aerodynamic characteristics of high-speed pantographs.A component optimization method is proposed as a possible solution to the problemof aerodynamic drag and noise in high-speed pantographs.The results of the study indicate that the panhead,base and insulator are the main contributors to aerodynamic drag and noise in high-speed pantographs.Therefore,a gradual optimization process is implemented to improve the most significant components that cause aerodynamic drag and noise.By optimizing the cross-sectional shape of the strips and insulators,the drag and noise caused by airflow separation and vortex shedding can be reduced.The aerodynamic drag of insulator with circular cross section and strips with rectangular cross section is the largest.Ellipsifying insulators and optimizing the chamfer angle and height of the windward surface of the strips can improve the aerodynamic performance of the pantograph.In addition,the streamlined fairing attached to the base can eliminate the complex flow and shield the radiated noise.In contrast to the original pantograph design,the improved pantograph shows a 21.1%reduction in aerodynamic drag and a 1.65 dBA reduction in aerodynamic noise.

Reinforcement Learning Based Quantization Strategy Optimal Assignment Algorithm for Mixed Precision

The quantization algorithm compresses the original network by reducing the numerical bit width of the model,which improves the computation speed. Because different layers have different redundancy and sensitivity to databit width. Reducing the data bit width will result in a loss of accuracy. Therefore, it is difficult to determinethe optimal bit width for different parts of the network with guaranteed accuracy. Mixed precision quantizationcan effectively reduce the amount of computation while keeping the model accuracy basically unchanged. In thispaper, a hardware-aware mixed precision quantization strategy optimal assignment algorithm adapted to low bitwidth is proposed, and reinforcement learning is used to automatically predict the mixed precision that meets theconstraints of hardware resources. In the state-space design, the standard deviation of weights is used to measurethe distribution difference of data, the execution speed feedback of simulated neural network accelerator inferenceis used as the environment to limit the action space of the agent, and the accuracy of the quantization model afterretraining is used as the reward function to guide the agent to carry out deep reinforcement learning training. Theexperimental results show that the proposed method obtains a suitable model layer-by-layer quantization strategyunder the condition that the computational resources are satisfied, and themodel accuracy is effectively improved.The proposed method has strong intelligence and certain universality and has strong application potential in thefield of mixed precision quantization and embedded neural network model deployment.

In situ calibrated angle between the quantization axis and the propagating direction of the light field for trapping neutral atoms

The recently developed magic-intensity trapping technique of neutral atoms efficiently mitigates the detrimental effect of light shifts on atomic qubits and substantially enhances the coherence time. This technique relies on applying a bias magnetic field precisely parallel to the wave vector of a circularly polarized trapping laser field. However, due to the presence of the vector light shift experienced by the trapped atoms, it is challenging to precisely define a parallel magnetic field, especially at a low bias magnetic field strength, for the magic-intensity trapping of85Rb qubits. In this work, we present a method to calibrate the angle between the bias magnetic field and the trapping laser field with the compensating magnetic fields in the other two directions orthogonal to the bias magnetic field direction. Experimentally, with a constantdepth trap and a fixed bias magnetic field, we measure the respective resonant frequencies of the atomic qubits in a linearly polarized trap and a circularly polarized one via the conventional microwave Rabi spectra with different compensating magnetic fields and obtain the corresponding total magnetic fields via the respective resonant frequencies using the Breit–Rabi formula. With known total magnetic fields, the angle is a function of the other two compensating magnetic fields.Finally, the projection value of the angle on either of the directions orthogonal to the bias magnetic field direction can be reduced to 0(4)° by applying specific compensating magnetic fields. The measurement error is mainly attributed to the fluctuation of atomic temperature. Moreover, it also demonstrates that, even for a small angle, the effect is strong enough to cause large decoherence of Rabi oscillation in a magic-intensity trap. Although the compensation method demonstrated here is explored for the magic-intensity trapping technique, it can be applied to a variety of similar precision measurements with trapped neutral atoms.

A Novel Quantization and Model Compression Approach for Hardware Accelerators in Edge Computing

Massive computational complexity and memory requirement of artificial intelligence models impede their deploy-ability on edge computing devices of the Internet of Things(IoT).While Power-of-Two(PoT)quantization is pro-posed to improve the efficiency for edge inference of Deep Neural Networks(DNNs),existing PoT schemes require a huge amount of bit-wise manipulation and have large memory overhead,and their efficiency is bounded by the bottleneck of computation latency and memory footprint.To tackle this challenge,we present an efficient inference approach on the basis of PoT quantization and model compression.An integer-only scalar PoT quantization(IOS-PoT)is designed jointly with a distribution loss regularizer,wherein the regularizer minimizes quantization errors and training disturbances.Additionally,two-stage model compression is developed to effectively reduce memory requirement,and alleviate bandwidth usage in communications of networked heterogenous learning systems.The product look-up table(P-LUT)inference scheme is leveraged to replace bit-shifting with only indexing and addition operations for achieving low-latency computation and implementing efficient edge accelerators.Finally,comprehensive experiments on Residual Networks(ResNets)and efficient architectures with Canadian Institute for Advanced Research(CIFAR),ImageNet,and Real-world Affective Faces Database(RAF-DB)datasets,indicate that our approach achieves 2×∼10×improvement in the reduction of both weight size and computation cost in comparison to state-of-the-art methods.A P-LUT accelerator prototype is implemented on the Xilinx KV260 Field Programmable Gate Array(FPGA)platform for accelerating convolution operations,with performance results showing that P-LUT reduces memory footprint by 1.45×,achieves more than 3×power efficiency and 2×resource efficiency,compared to the conventional bit-shifting scheme.

Image Steganography by Pixel-Value Differencing Using General Quantization Ranges

A new steganographic method by pixel-value differencing(PVD)using general quantization ranges of pixel pairs’difference values is proposed.The objective of this method is to provide a data embedding technique with a range table with range widths not limited to powers of 2,extending PVD-based methods to enhance their flexibility and data-embedding rates without changing their capabilities to resist security attacks.Specifically,the conventional PVD technique partitions a grayscale image into 1×2 non-overlapping blocks.The entire range[0,255]of all possible absolute values of the pixel pairs’grayscale differences in the blocks is divided into multiple quantization ranges.The width of each quantization range is a power of two to facilitate the direct embedding of the bit information with high embedding rates.Without using power-of-two range widths,the embedding rates can drop using conventional embedding techniques.In contrast,the proposed method uses general quantization range widths,and a multiple-based number conversion mechanism is employed skillfully to implement the use of nonpower-of-two range widths,with each pixel pair being employed to embed a digit in the multiple-based number.All the message bits are converted into a big multiple-based number whose digits can be embedded into the pixel pairs with a higher embedding rate.Good experimental results showed the feasibility of the proposed method and its resistance to security attacks.In addition,implementation examples are provided,where the proposed method adopts non-power-of-two range widths and employsmultiple-based number conversion to expand the data-hiding and steganalysis-resisting capabilities of other PVD methods.

Vibration and noise mechanism of a 110 kV transformer under DC bias based on finite element method

Global energy and environmental issues are becoming increasingly problematic,and the vibration and noise problem of 110 kV transformers,which are the most widely distributed,have attracted widespread attention from both inside and outside the industry.DC bias is one of the main contributing factors to vibration noise during the normal operation of transformers.To clarify the vibration and noise mechanism of a 110 kV transformer under a DC bias,a multi-field coupling model of a 110 kV transformer was established using the finite element method.The electromagnetic,vibration,and noise characteristics during the DC bias process were compared and quantified through field circuit coupling in parallel with the power frequency of AC,harmonic,and DC power sources.It was found that a DC bias can cause significant distortions in the magnetic flux density,force,and displacement distributions of the core and winding.The contributions of the DC bias effect to the core and winding are different at Kdc=0.85.At this point,the core approached saturation,and the increase in the core force and displacement slowed.However,the saturation of the core increased the leakage flux,and the stress and displacement of the winding increased faster.The sound field distribution characteristics of the 110 kV transformer under a DC bias are related to the force characteristics.When the DC bias coefficient was 1.25,the noise sound pressure level reached 73.6 dB.

Learning Vector Quantization-Based Fuzzy Rules Oversampling Method

Imbalanced datasets are common in practical applications,and oversampling methods using fuzzy rules have been shown to enhance the classification performance of imbalanced data by taking into account the relationship between data attributes.However,the creation of fuzzy rules typically depends on expert knowledge,which may not fully leverage the label information in training data and may be subjective.To address this issue,a novel fuzzy rule oversampling approach is developed based on the learning vector quantization(LVQ)algorithm.In this method,the label information of the training data is utilized to determine the antecedent part of If-Then fuzzy rules by dynamically dividing attribute intervals using LVQ.Subsequently,fuzzy rules are generated and adjusted to calculate rule weights.The number of new samples to be synthesized for each rule is then computed,and samples from the minority class are synthesized based on the newly generated fuzzy rules.This results in the establishment of a fuzzy rule oversampling method based on LVQ.To evaluate the effectiveness of this method,comparative experiments are conducted on 12 publicly available imbalance datasets with five other sampling techniques in combination with the support function machine.The experimental results demonstrate that the proposed method can significantly enhance the classification algorithm across seven performance indicators,including a boost of 2.15%to 12.34%in Accuracy,6.11%to 27.06%in G-mean,and 4.69%to 18.78%in AUC.These show that the proposed method is capable of more efficiently improving the classification performance of imbalanced data.

Adaptive Bistable Stochastic Resonance Based Weak Signal Reception in Additive Laplacian Noise

Weak signal reception is a very important and challenging problem for communication systems especially in the presence of non-Gaussian noise,and in which case the performance of optimal linear correlated receiver degrades dramatically.Aiming at this,a novel uncorrelated reception scheme based on adaptive bistable stochastic resonance(ABSR)for a weak signal in additive Laplacian noise is investigated.By analyzing the key issue that the quantitative cooperative resonance matching relationship between the characteristics of the noisy signal and the nonlinear bistable system,an analytical expression of the bistable system parameters is derived.On this basis,by means of bistable system parameters self-adaptive adjustment,the counterintuitive stochastic resonance(SR)phenomenon can be easily generated at which the random noise is changed into a benefit to assist signal transmission.Finally,it is demonstrated that approximately 8dB bit error ratio(BER)performance improvement for the ABSR-based uncorrelated receiver when compared with the traditional uncorrelated receiver at low signal to noise ratio(SNR)conditions varying from-30dB to-5dB.

Can urban forests provide acoustic refuges for birds?Investigating the influence of vegetation structure and anthropogenic noise on bird sound diversity

As a crucial component of terrestrial ecosystems,urban forests play a pivotal role in protecting urban biodiversity by providing suitable habitats for acoustic spaces.Previous studies note that vegetation structure is a key factor influencing bird sounds in urban forests;hence,adjusting the frequency composition may be a strategy for birds to avoid anthropogenic noise to mask their songs.However,it is unknown whether the response mechanisms of bird vocalizations to vegetation structure remain consistent despite being impacted by anthropogenic noise.It was hypothesized that anthropogenic noise in urban forests occupies the low-frequency space of bird songs,leading to a possible reshaping of the acoustic niches of forests,and the vegetation structure of urban forests is the critical factor that shapes the acoustic space for bird vocalization.Passive acoustic monitoring in various urban forests was used to monitor natural and anthropogenic noises,and sounds were classified into three acoustic scenes(bird sounds,human sounds,and bird-human sounds)to determine interconnections between bird sounds,anthropogenic noise,and vegetation structure.Anthropogenic noise altered the acoustic niche of urban forests by intruding into the low-frequency space used by birds,and vegetation structures related to volume(trunk volume and branch volume)and density(number of branches and leaf area index)significantly impact the diversity of bird sounds.Our findings indicate that the response to low and high frequency signals to vegetation structure is distinct.By clarifying this relationship,our results contribute to understanding of how vegetation structure influences bird sounds in urban forests impacted by anthropogenic noise.

A Denoiser for Correlated Noise Channel Decoding: Gated-Neural Network

This letter proposes a sliced-gated-convolutional neural network with belief propagation(SGCNN-BP) architecture for decoding long codes under correlated noise. The basic idea of SGCNNBP is using Neural Networks(NN) to transform the correlated noise into white noise, setting up the optimal condition for a standard BP decoder that takes the output from the NN. A gate-controlled neuron is used to regulate information flow and an optional operation—slicing is adopted to reduce parameters and lower training complexity. Simulation results show that SGCNN-BP has much better performance(with the largest gap being 5dB improvement) than a single BP decoder and achieves a nearly 1dB improvement compared to Fully Convolutional Networks(FCN).

Dynamics and synchronization in a memristor-coupled discrete heterogeneous neuron network considering noise

Research on discrete memristor-based neural networks has received much attention.However,current research mainly focuses on memristor–based discrete homogeneous neuron networks,while memristor-coupled discrete heterogeneous neuron networks are rarely reported.In this study,a new four-stable discrete locally active memristor is proposed and its nonvolatile and locally active properties are verified by its power-off plot and DC V–I diagram.Based on two-dimensional(2D)discrete Izhikevich neuron and 2D discrete Chialvo neuron,a heterogeneous discrete neuron network is constructed by using the proposed discrete memristor as a coupling synapse connecting the two heterogeneous neurons.Considering the coupling strength as the control parameter,chaotic firing,periodic firing,and hyperchaotic firing patterns are revealed.In particular,multiple coexisting firing patterns are observed,which are induced by different initial values of the memristor.Phase synchronization between the two heterogeneous neurons is discussed and it is found that they can achieve perfect synchronous at large coupling strength.Furthermore,the effect of Gaussian white noise on synchronization behaviors is also explored.We demonstrate that the presence of noise not only leads to the transition of firing patterns,but also achieves the phase synchronization between two heterogeneous neurons under low coupling strength.

Research on the generation method of seawater sound velocity model based on Perlin noise

In the processing of conventional marine seismic data,seawater is often assumed to have a constant velocity model.However,due to static pressure,temperature difference and other factors,random disturbances may often frequently in seawater bodies.The impact of such disturbances on data processing results is a topic of theoretical research.Since seawater sound velocity is a difficult physical quantity to measure,there is a need for a method that can generate models conforming to seawater characteristics.This article will combine the Munk model and Perlin noise to propose a two-dimensional dynamic seawater sound velocity model generation method,a method that can generate a dynamic,continuous,random seawater sound velocity model with some regularity at large scales.Moreover,the paper discusses the influence of the inhomogeneity characteristics of seawater on wave field propagation and imaging.The results show that the seawater sound velocity model with random disturbance will have a significant influence on the wave field simulation and imaging results.

Hierarchical Controller Synthesis Under Linear Temporal Logic Specifications Using Dynamic Quantization

Linear temporal logic(LTL)is an intuitive and expressive language to specify complex control tasks,and how to design an efficient control strategy for LTL specification is still a challenge.In this paper,we implement the dynamic quantization technique to propose a novel hierarchical control strategy for nonlinear control systems under LTL specifications.Based on the regions of interest involved in the LTL formula,an accepting path is derived first to provide a high-level solution for the controller synthesis problem.Second,we develop a dynamic quantization based approach to verify the realization of the accepting path.The realization verification results in the necessity of the controller design and a sequence of quantization regions for the controller design.Third,the techniques of dynamic quantization and abstraction-based control are combined together to establish the local-to-global control strategy.Both abstraction construction and controller design are local and dynamic,thereby resulting in the potential reduction of the computational complexity.Since each quantization region can be considered locally and individually,the proposed hierarchical mechanism is more efficient and can solve much larger problems than many existing methods.Finally,the proposed control strategy is illustrated via two examples from the path planning and tracking problems of mobile robots.

Effect of Bogie Cavity End Wall Inclination on Flow Field and Aerodynamic Noise in the Bogie Region of High-Speed Trains

Combining the detached eddy simulation(DES)method and Ffowcs Williams-Hawkings(FW-H)equation,the effect of bogie cavity end wall inclination on the flow field and aerodynamic noise in the bogie region is numerically studied.First,the simulation is conducted based on a simplified cavity-bogie model,including five cases with different inclination angles of the front and rear walls of the cavity.By comparing and analyzing the flow field and acoustic results of the five cases,the influence of the regularity and mechanism of the bogie cavity end wall inclination on the flow field and the aerodynamic noise of the bogie region are revealed.Then,the noise reduction strategy determined by the results of the simplified cavity-bogie model is applied to a three-car marshaling train model to verify its effectiveness when applied to the real train.The results reveal that the forward inclination of the cavity front wall enlarges the influence area of shear vortex structures formed at the leading edge of the cavity and intensifies the interaction between the vortex structures and the front wheelset,frontmotor,and front gearbox,resulting in the increase of the aerodynamic noise generated by the bogie itself.The backward inclination of the cavity rear wall is conducive to guiding the vortex structures flow out of the cavity and weakening the interaction between the shear vortex structures and the cavity rear wall,leading to the reduction of the aerodynamic noise generated by the bogie cavity.Inclining the rear end wall of the foremost bogie cavity of the head car is a feasible aerodynamic noise reduction measure for high-speed trains.

Research on the flow stability and noise reduction characteristics of quasi-periodic elastic support skin

To enhance flow stability and reduce hydrodynamic noise caused by fluctuating pressure,a quasiperiodic elastic support skin composed of flexible walls and elastic support elements is proposed for fluid noise reduction.The arrangement of the elastic support element is determined by the equivalent periodic distance and quasi-periodic coefficient.In this paper,a dynamic model of skin in a fluid environment is established.The influence of equivalent periodic distance and quasi-periodic coefficient on flow stability is investigated.The results suggest that arranging the elastic support elements in accordance with the quasi-periodic law can effectively enhance flow stability.Meanwhile,the hydrodynamic noise calculation results demonstrate that the skin exhibits excellent noise reduction performance,with reductions of 10 dB in the streamwise direction,11 dB in the spanwise direction,and 10 dB in the normal direction.The results also demonstrate that the stability analysis method can serve as a diagnostic tool for flow fields and guide the design of noise reduction structures.

Physical Layer Encryption of OFDM-PON Based on Quantum Noise Stream Cipher with Polar Code

Orthogonal frequency division multiplexing passive optical network(OFDM-PON) has superior anti-dispersion property to operate in the C-band of fiber for increased optical power budget. However,the downlink broadcast exposes the physical layer vulnerable to the threat of illegal eavesdropping. Quantum noise stream cipher(QNSC) is a classic physical layer encryption method and well compatible with the OFDM-PON. Meanwhile, it is indispensable to exploit forward error correction(FEC) to control errors in data transmission. However, when QNSC and FEC are jointly coded, the redundant information becomes heavier and thus the code rate of the transmitted signal will be largely reduced. In this work, we propose a physical layer encryption scheme based on polar-code-assisted QNSC. In order to improve the code rate and security of the transmitted signal, we exploit chaotic sequences to yield the redundant bits and utilize the redundant information of the polar code to generate the higher-order encrypted signal in the QNSC scheme with the operation of the interleaver.We experimentally demonstrate the encrypted 16/64-QAM, 16/256-QAM, 16/1024-QAM, 16/4096-QAM QNSC signals transmitted over 30-km standard single mode fiber. For the transmitted 16/4096-QAM QNSC signal, compared with the conventional QNSC method, the proposed method increases the code rate from 0.1 to 0.32 with enhanced security.

Research on the Generation Mechanism and Suppression Method of Aerodynamic Noise in Expansion Cavity Based on Hybrid Method

The expansion chamber serves as the primary silencing structure within the exhaust pipeline.However,it can also act as a sound-emitting structure when subjected to airflow.This article presents a hybrid method for numerically simulating and analyzing the unsteady flow and aerodynamic noise in an expansion chamber under the influence of airflow.A fluid simulation model is established,utilizing the Large Eddy Simulation(LES)method to calculate the unsteady flow within the expansion chamber.The simulation results effectively capture the development and changes of the unsteady flow and vorticity inside the cavity,exhibiting a high level of consistency with experimental observations.To calculate the aerodynamic noise sources within the cavity,the flow field results are integrated using the method of integral interpolation and inserted into the acoustic grid.The acoustic analogy method is then employed to determine the aerodynamic noise sources.An acoustic simulation model is established,and the flow noise source is imported into the sound field grid to calculate the sound pressure at the far-field response point.The calculated sound pressure levels and resonance frequencies show good agreement with the experimental results.To address the issue of airflow regeneration noise within the cavity,perforated tubes are selected as a means of noise suppression.An experimental platformfor airflow regeneration noise is constructed,and experimental samples are processed to analyze and verify the noise suppression effect of perforated tube expansion cavities under different airflow velocities.The research findings indicate that the perforated tube expansion cavity can effectively suppress low-frequency aerodynamic noise within the cavity by impeding the formation of strong shear layers.Moreover,the semi-perforated tube expansion cavity demonstrates the most effective suppression of aerodynamic noise.

Numerical modeling techniques for noise emission of free railway wheels

In this article,we consider the numerical prediction of the noise emission from a wheelset in laboratory conditions.We focus on the fluid-structure interaction leading to sound emission in the fluid domain by analyzing three different methods to account for acoustic sources.These are a discretized baffled piston using the discrete calculation method(DCM),a closed cylindrical volume using the boundary element method(BEM)and radiating elastic disks in a cubic enclosure solved with the finite element method(FEM).We provide the validation of the baffled piston and the BEM using measurements of the noise emission of a railway wheel by considering ground reflections in the numerical models.Selected space-resolved waveforms are compared with experimental results as well as with a fluid-structure interaction finite element model.The computational advantage of a discretized disk mounted on a baffle and BEM compared to FEM is highlighted,and the baffled pistons limitations caused by a lack of edge radiation effects are investigated.

Symmetric Brownian motor subjected to Lévy noise

In the past few years,attention has mainly been focused on the symmetric Brownian motor(BM)with Gaussian noises,whose current and energy conversion efficiency are very low.Here,we investigate the operating performance of the symmetric BM subjected to Lévy noise.Through numerical simulations,it is found that the operating performance of the motor can be greatly improved in asymmetric Lévy noise.Without any load,the Lévy noises with smaller stable indexes can let the motor give rise to a much greater current.With a load,the energy conversion efficiency of the motor can be enhanced by adjusting the stable indexes of the Lévy noises with symmetry breaking.The results of this research are of great significance for opening up BM’s intrinsic physical mechanism and promoting the development of nanotechnology.