Filter Dimensions Research Articles

APSSF: Adaptive CNN Pruning Based on Structural Similarity of Filters

Convolutional neural network (CNN) pruning is a technique used to remove redundant parameters from the network. By doing so, it aims to greatly reduce the computational complexity and scale of the network while still preserving its accuracy. In the CNN, the majority of parameters are weights that form filters. When it comes to pruning, it is more effective to focus on removing redundant filters rather than insignificant weights within filters. The essence of filter pruning lies in determining the significance or contribution of each filter. Filters that have a significant contribution are kept, while others are pruned. Current methods for calculating contribution in pruning often rely on weight magnitude or filter similarity. However, approaches based solely on assume that small weights are unimportant and ignore correlation between filters, which leads to a significant loss of network accuracy. Those based on filter similarity flatten filter tensors into a vector when calculating filter similarity, and lose the important structural information of filters, or the superposition information of the weight convolution in the corresponding space position. These limitations can compromise the accuracy and effectiveness of the pruning process. This paper proposes an adaptive CNN pruning method based on the structural similarity of filters (APSSF) by taking both the structural characteristics of and the correlation between filters into the consideration for pruning filters. APSSF efficiently calculates the distance between the filters by factoring in information from all the dimensions of filters, and clusters the filters according to the distance threshold determined adaptively according to the compression rate, and deletes a certain number of filters from each category. On the CIFAR10 and ImageNet datasets, APSSF outperforms several state-of-the-art methods. On the CIFAR100, APSSF reduces parameters of networks by 91.71% and 74.80% on VGG-16 and ResNet-34, respectively. The accuracy was decreased only by 0.03 on VGG-16, while on ResNet-34, it was increased by 0.04.

Exploring the impact of oil security attention on oil volatility: A new perspective

AbstractBy constructing a novel index, the oil security attention index, this paper uses the heterogeneous autoregressi (HAR)‐type and its extended models to study whether oil security attention can predict oil volatility. Based on the definition of the different dimensions of oil security and three‐pass regression filter (TPRF) dimension reduction technology, combined with Google search volume data of 23 keywords related to oil security, the oil security attention index is constructed. Considering the potential nonlinear relationship between attention and oil volatility, we incorporate asymmetric effects in the new extended HAR‐type models. The research findings show that the oil security attention index we propose can capture the volatility of West Texas Intermediate. The out‐of‐sample results indicate that the extended models have better predictive power, which confirms the asymmetric relationship between oil security attention and oil volatility. In the robustness analysis, we compare TPRF with traditional principal component analysis (PCA) and partial least squares (PLS), and show that the oil security attention index constructed using TPRF has more favourable information than PCA and PLS to capture the oil security attention of the public.

TinyAD: Memory-Efficient Anomaly Detection for Time-Series Data in Industrial IoT

Monitoring and detecting abnormal events in cyber-physical systems is crucial to industrial production. With the prevalent deployment of the Industrial Internet of Things (IIoT), an enormous amount of time series data is collected to facilitate machine learning models for anomaly detection, and it is of the utmost importance to directly deploy the trained models on the IIoT devices. However, it is most challenging to deploy complex deep learning models such as Convolutional Neural Networks (CNNs) on these memory-constrained IIoT devices embedded with microcontrollers (MCUs). To alleviate the memory constraints of MCUs, we propose a novel framework named Tiny Anomaly Detection (TinyAD) to efficiently facilitate onboard inference of CNNs for real-time anomaly detection. First, we conduct a comprehensive analysis of depthwise separable CNNs and regular CNNs for anomaly detection and find that the depthwise separable convolution operation can reduce the model size by 50-90% compared with the traditional CNNs. Then, to reduce the peak memory consumption of CNNs, we explore two complementary strategies, in-place, and patch-by-patch memory rescheduling, and integrate them into a unified framework. The in-place method decreases the peak memory of the depthwise convolution by sparing a temporary buffer to transfer the activation results, while the patch-by-patch method further reduces the peak memory of layer-wise execution by slicing the input data into corresponding receptive fields and executing in order. Furthermore, by adjusting the dimension of convolution filters, these strategies apply to both univariate time series and multidomain time series features. Extensive experiments on real-world industrial datasets show that our framework can reduce peak memory consumption by 2-5x with negligible computation overhead.

Removal of Emerging Contaminants by Degradation during Filtration: A Review of Experimental Procedures and Modeling

Here, we review the efficient removal of organic micropollutants from water by degradation during filtration using specialized bacteria and enzymes. In both approaches, the filter provides essential binding sites where efficient degradation can occur. A model is presented that enables the simulation and prediction of the kinetics of filtration for a given pollutant concentration, flow rate, and filter dimensions and can facilitate the design of experiments and capacity estimates; it predicts the establishment of a steady state, during which the emerging concentrations of the pollutants remain constant. One method to remove cyanotoxins produced by Microcystis cyanobacteria, which pose a threat at concentrations above 1.0 µg L−1, is to use an activated granular carbon filter with a biofilm; this method resulted in the complete removal of the filtered toxins (5 µg L−1) during a long experiment (225 d). This system was analyzed using a model which predicted complete toxin removal when applied at a 10-fold-higher concentration. Enzymes are also used in filtration processes for the degradation of trace organic contaminants, mostly through the use of membrane bioreactors, where the enzyme is continuously introduced or maintained in the bioreactor, or it is immobilized on the membrane.

Design of a Compact Microstrip Filtenna for Miniaturized Devices to Access Internet of Things Using Long Term Evolution

A filtenna or filtering antenna has both radiation and filtering functions and is an important module for the RF front-end of wireless devices. The main function of a filtering antenna is to filter out the out-of-band interferences from the adjacent bands. In this paper, the design of a compact coplanar microstrip filtering antenna at 2.6GHz band is presented. The antenna is designed for the miniaturized devices for the city connectivity to access the internet of things (IoT) services over 2.6GHz long term evolution (LTE) gateways. Filtering antenna is designed using hairpin bandpass filter integrated with semi-circular microstrip antenna. Filtering antenna is designed by computer simulation technology (CST) software and measurements are performed by vector network analyzer. Good impedance matching at 2.6GHz band is obtained for the filtering antenna by choosing appropriate dimensions of the hairpin bandpass filter. The simulated 10dB return loss bandwidth of filtering antenna is 4.65% with maximum gain of 1.9dB. Measured and simulated results agree well for the semi-circular filtenna.

Design of a Microstrip Filtering Antenna for 4G and 5G Wireless Networks

The filtering antenna provides both radiation and filtering features and is an important component for the RF front-end of wireless devices. The main function of a filtering antenna is to reject out-of-band signals, thus reducing the interference from adjacent channels. The aim of the present work is to design a 2.6 GHz microstrip filtering antenna for 4G and 5G global mobile services. The filtering antenna is designed using a hairpin bandpass filter integrated with an elliptical microstrip aerial. Good impedance matching is obtained by using appropriate dimensions of the hairpin bandpass filter. The 10 dB return loss bandwidth of the filtering antenna is approx. 5.7%, with the maximum gain for the elliptical filtering antenna of approx. 2.2 dB. Good agreements between the measured and simulated results are obtained for the proposed filtering antenna and the bandwidth covers almost the entire 2.6 GHz band.

Efficient segment-update block LMS-Newton algorithm for active control of road noise

For active control of road noise inside a vehicle, many reference signals are required in adaptive control systems, resulting in slow convergence speeds and heavy computational loads. In this paper, a means of reducing the computational complexity of adaptive control algorithms is investigated based on the fast-converging least-mean-squares Newton (LMS-Newton) algorithm. By analyzing the autocorrelation matrix of reference signals filtered by secondary paths, it is found that the dimensions of the autocorrelation matrix of filtered reference signals do not necessarily have to equal the dimensions of the adaptive control filters, and instead the dimensions can be determined from the correlation between reference signals filtered by secondary paths. Then a computationally efficient segment-update block algorithm is proposed that splits adaptive control filters into several segments and the coefficients of each segment are updated independently. Theoretical analysis shows that the proposed segment-update block algorithm has similar convergence speed and noise-reduction performance as the traditional LMS-Newton algorithm when a segment of appropriate length is chosen, but the computational complexity is reduced dramatically because the inverse autocorrelation matrix is replaced by a smaller matrix. The performance of the proposed algorithm is verified through simulation experiments with different recorded road noises and secondary paths, and the results are compared with those of some typical algorithms. The simulation results showed that the proposed algorithm achieved a similar noise-reduction performance with an approximately 8.9-times reduction of computational complexity compared with the traditional LMS-Newton algorithm.

Single-band Series Absorptive Common-mode Noise Filter

A Single-band Series Absorptive Common Mode Noise Filter (ACMF) is proposed. The ACMF is embedded in a four-layer printed circuit board (PCB) and consists of three parts: a Reflective Common Mode Noise Filter (RCMF), a matching circuit, and an absorber. The RCMF is designed using mushroom-type resonators. The matching circuit is designed using meander lines to reduce the size of the filter dimensions. The absorber of the Common Mode noise (CM) is a series resistor. The designed operating frequency is 2.45 GHz. The simulation results are as follows: the insertion loss of CM (Scc21) is -22.49 dB at the frequency of 2.61 GHz, the return loss of CM (Scc21) is -18.62 dB at the frequency of 2.5 GHz, while the integrity of the Differential Mode signals (DM) can be maintained with a very small insertion loss (Sdd21) of -1 dB at the frequency range of 0-8 GHz, and the achieved Absorption Efficiency (AE) is 93% at the frequency of 2.54 GHz. The proposed ACMF dimension is 10.3 x4.6 mm. The fractional bandwidth is 19%. The measurement results of the fabricated ACMF do not deviate significantly the simulation results. They are as follows: Scc21 is -17.87 dB at the frequency of 2.31 GHz, Scc11 is -20.87 dB at the frequency of 2.38 GHz, Sdd21 is -2.8 dB at the frequency range of 0–8 GHz, the Absorption Efficiency is 97% at the frequency of 2.32 GHz, and the fractional bandwidth is 17%. Therefore, the results of the ACMF design carried out by simulation can be implemented into a fabricated ACMF with measurement results similar to the calculation results in the design.

A Miniaturized Dual-Band Diplexer Design with High Port Isolation for UHF/SHF Applications Using a Neural Network Model

In this paper, a compact dual-band diplexer is proposed using two interdigital filters. The proposed microstrip diplexer correctly works at 2.1 GHz and 5.1 GHz. In the proposed diplexer, two fifth-order bandpass interdigital filters are designed to pass the desired frequency bands. Applied interdigital filters with simple structures pass the 2.1 GHz and 5.1 GHz frequencies and suppress other frequency bands with high attenuation levels. The dimensions of the interdigital filter are obtained using the artificial neural network (ANN) model, constructed from the EM-simulation data. The desired filter and diplexer parameters, such as operating frequency, bandwidth, and insertion loss, can be obtained using the proposed ANN model. The insertion loss parameter of the proposed diplexer is 0.4 dB, and more than 40 dB output port isolation is obtained (for both operating frequencies). The main circuit has the small size of 28.5 mm × 23 mm (0.32 λg × 0.26 λg). The proposed diplexer, with the achieved desired parameters, is a good candidate for UHF/SHF applications.

A Novel Stacked Ceramic-Monoblock Filter With Highly Flexible Transmission Zero Pair

In this letter, a novel five-pole monoblock dielectricresonator (MDR) bandpass filter with two transmission zeroes (TZs) based on the compact stacked configuration is proposed. The TZ pair is introduced with great flexibility and can be readily allocated at either or both sides of the passband with only a few filter dimension adjustments. The proposed filter has a simple fabrication process with tuning flexibility since there is no need to weld or assemble cavities. A 2.6-GHz filter with a symmetric TZ pair and 150-MHz bandwidth is fabricated. Measurement results show good agreement with simulated responses for validating the design theory.

Ku-Band SIW Filter with High Fractional Bandwidth Optimized Using Feed Forward Back Propagation ANN

An Artificial Neural Network (ANN) based Substrate Integrated Waveguide Band Pass Filter (SIW BPF) is proposed in this paper. A Feed Forward Back Propagation Neural Network (FF-BP NN) is utilized to optimize the filter dimensions. The Gradient Descent with Momentum and Adaptive Learning Rate algorithm is used to train the network at a learning rate of 0.01. The high pass characteristics of the SIW is converted into band pass by incorporating U slots on its top plane. Defected Ground Structure (DGS) is utilized in the bottom plane to improve the impedance matching. To validate, the prototype is fabricated using RT/Duroid 5880 and tested. The proposed filter has a center frequency at 14.68 GHz with a wide pass band from 12.92 to 16.43 GHz with a 3-dB Fractional Bandwidth (FBW) of 24%, return loss more than 20 dB and insertion loss of about 1.9 dB within the pass band. The filter has a small dimension of about 0.63 λ g 2 , where λ g is the guided wavelength at the center frequency. This filter offers wide passband, smaller size, low insertion loss, good return loss and it is useful in Ku-band satellite communication applications.

A Novel Filter With High Harmonics Attenuation and Small Dimension for Grid-Connected Inverter

Passive power filters are indispensable for grid-connected inverters (GCIs) to eliminate current harmonics. However, typical filters require large inductance, which not only increases the filter dimension and costs but also leads to slower current response and a high voltage drop across the filter. This article proposes a novel filter called LCTCL filter, which effectively attenuates harmonics with small inductance. In addition, this article presents a controller synthesis to the GCI with LCTCL filter without an additional resistor. We demonstrate the effectiveness of the proposed LCTCL filter through comparative analysis with typical filters via simulations and experiments. We clarify that the proposed LCTCL filter has excellent features in terms of filter dimension, frequency characteristics, robustness for parameter variations, harmonic suppression, and total harmonic distortion.

Transmission-Line Schematic for Harmonic-Controlled Bandpass Filters

This article presents a rigorous synthesis method of a transmission-line filter whose second passband can be placed at any multiple of the center frequency, in theory. Two coupled-line schematics for harmonic-controlled bandpass filters are derived from fundamental <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">LC</i> filter prototypes. Their design parameters are provided in closed-form expressions so filter dimensions can be acquired rigorously without relying on EM-simulation-based coupling tests. Three filters were designed and fabricated for verification purposes, and their measurement results exhibited excellent agreement with the theory.

Convolution-layer parameters optimization in Convolutional Neural Networks

Convolutional Neural Networks (CNNs) are the most important deep learning algorithms to classify images based on their visual features. CNNs architectures are made of convolution, pooling and fully-connected (FC) layers and their corresponding parameters, which affect classification performance significantly. Convolution parameters optimization for CNNs, referred as CPOCNN, is proposed in this paper. To the best of our knowledge, this is the first optimization model which assigns adaptive upper-bounds of convolution parameters depends on data dimension in current layer and number of remained layers to reach the output layer. For this task, a comprehensive mathematical model is presented for constant structures of CNNs and it is proven that more optimization space is explored in comparison with all state-of-the-art methods. In optimization process, number of convolution filters and type of pooling filters are selected randomly; dimension of pooling filters, zero-padding and stride are considered constant. CPOCNN has been evaluated on 7 publicly available datasets and compared with 53 competitive CNN models with constant and optimized structures. Interested results showed that CPOCNN not only outperforms state-of-the-art CNN methods but also enriches weak CNN models by improving their accuracies more than 35%. Source Code of this paper is available online at Github.11https://github.com/kohzadi/CHOCNN.git.

A Selective Wideband Bandpass Filter With Wide Stopband Using Mixed Lumped-Distributed Circuits

In this brief, a wideband bandpass filter with extremely wide spurious-free stopband and high selectivity is presented. The filter consists of a wideband bandpass cell cascaded with sharp-rejection wide-stopband low-pass filters (WSLPFs), both designed using a mix of lumped and distributed elements. The proposed wideband bandpass cell, consisting of two series LC resonators and shunt-connected open-ended microstrip stubs, produces a three-pole passband with two transmission zeros on its opposite sides. The LC resonators provide suppression of spurious transmission peaks arising due to the main stub, which is further enhanced by the auxiliary stubs in the wideband cell and the WSLPFs. The design technique imbibes the effects of the self-resonating frequencies of the lumped components, and thus enable the filter to achieve the stopband extension irrespective of their size and make. A prototype of the proposed filter, designed at 1.25 GHz with 62.4% bandwidth, shows the stopband extended up to 18.6 times the center frequency with minimum rejection of 19 dB. The selectivity factor of the passband, defined as the 3-dB to 30-dB bandwidth ratio, is 0.65. The maximum filter dimension is less than a quarter of the wavelength at the center frequency, providing a miniature design.

A New Miniature Micro-Strip Two-Layer Band-Pass Filter Using Aperture-Coupled Hairpin Resonators

The goal of this project was to provide novel band-pass filter design techniques for mobile communications, which allow a significant reduction in the size of the filters produced. The novelty comes from the transformation of the single layer technique into a double layer technique by inserting coupling slots in a common mass plane. Because of their tiny size, these filters are suitable candidates for integration into mobile communication systems. Indeed, when compared to the dimensions of a single-layer planar filter, the multilayer construction allowed us to reduce the size of the filter by more than 40%. Five U-shaped hairpin resonators were placed on two micro-strip layers in the planned filter. Two apertures etched on a common ground plane positioned between the two layers allow varied couplings between the upper and bottom layer resonators. A five-pole hairpin band-pass filter was created as a result.

A capacitor based single source MLI with natural balancing and less component for EV/HEV application

SummaryDue to their remarkable performance, capacitor‐based inverters have recently gained attention. Hence, a new capacitor‐based multilevel inverter is presented in this paper for electric and hybrid electric vehicle (EV and HEV) applications. EV systems are quite well for their use of two‐level inverters; however, the generated load voltage comprises substantial undesirable harmonic content. It is regarded as one of the most efficient methods since replacing a two‐level inverter with a multilevel inverter improves the power quality despite significantly reducing total harmonic distortion. Therefore, the recommended filter dimension will also be minimized. A flurry of reliability concerns has arisen due to the increased number of devices, circuit complexity, and stress on the circuit devices. A nine‐level voltage waveform is created with only ten IGBTs, a DC‐Source, and two capacitors. In the proposed nine‐level inverter, the capacitor voltage is balanced utilizing a simple control approach to regulate the flying capacitor (FC) voltages actively. Here described a simple logic gate‐based pulse‐width modulation technique that ensures capacitor power balancing. The proposed inverter operation and capability are validated by experimental results derived from a laboratory prototype. Finally, by contrasting the new and standard inverter topologies, the virtues of the suggested architecture by the number of devices and price of the equipment are highlighted, and it is a simpler structure that requires less space and footprint area.

Implementation and optimization of the PTOLEMY transverse drift electromagnetic filter

The PTOLEMY transverse drift filter is a new concept to enable precision analysis of the energy spectrum of electrons near the tritium β-decay endpoint. This paper details the implementation and optimization methods for successful operation of the filter for electrons with a known pitch angle. We present the first demonstrator that produces the required magnetic field properties with an iron return-flux magnet. Two methods for the setting of filter electrode voltages are detailed. The challenges of low-energy electron transport in cases of low field are discussed, such as the growth of the cyclotron radius with decreasing magnetic field, which puts a ceiling on filter performance relative to fixed filter dimensions. Additionally, low pitch angle trajectories are dominated by motion parallel to the magnetic field lines and introduce non-adiabatic conditions and curvature drift. To minimize these effects and maximize electron acceptance into the filter, we present a three-potential-well design to simultaneously drain the parallel and transverse kinetic energies throughout the length of the filter. These optimizations are shown, in simulation, to achieve low-energy electron transport from a 1 T iron core (or 3 T superconducting) starting field with initial kinetic energy of 18.6 keV drained to < 10 eV (< 1 eV) in about 80 cm. This result for low field operation paves the way for the first demonstrator of the PTOLEMY spectrometer for measurement of electrons near the tritium endpoint to be constructed at the Gran Sasso National Laboratory (LNGS) in Italy.

Compact Dual Bandstop Filter Using Folded Lines and Genetic Algorithm

A novel method for creating a dual band bandstop filter with a meandering architecture is presented. Proposed bandstop filter shows two notch bands, one at 5.6 to 7.2 GHz and the other at 8 to 11 GHz. The designed filters simulated on HFSS software and fabricated on FR-4 substrates, filter dimensions have been optimized using GA. Filter shows S11 of 1dB and S21 of -18 dB at notch frequency 6.4 GHz and S11 of 1.2dB and S21 of -38 dB at notch frequency 9 GHz. The suggested filter has two pass bands one is up to 5 GHz for wireless communication and another is from 7.4 to 7.9 GHz for C band applications.

Design and Tuning of Extracted-Pole Filters With Non-Resonant Nodes by Circuit Model Extraction

This article proposes an analytical circuit model extraction technique for extracted-pole filters with nonresonant nodes (NRNs). The circuit model is described by a generalized coupling matrix incorporating both resonant nodes and NRNs. It is shown that, in most cases, Y-parameter polynomials of extracted-pole filters with NRNs take the same form as those of coupled-resonator filters. Thus, the model-based vector fitting technique can be applied to obtain legitimate rational function approximation to measured or simulated S-parameter data, and then, a circuit model in the correct form can be synthesized accordingly. This article discusses different possible arrangements of extracted-pole sections to develop a general solution to the circuit model extraction problem. It also defines a standard circuit form to rule out multiple solutions in the circuit model extraction problem for guiding the computer-aided design and tuning process. A combline filter design and a waveguide filter tuning example are presented to demonstrate the application of the circuit model extraction technique. These examples show that simple mapping relationships exist between extracted circuit parameters and physical elements. Thus, accurate dimensioning of extracted-pole filters can be accomplished without electromagnetic optimization in the design phase. Filter fine-tuning can be done by inexperienced operators or an automatic tuning machine with the circuit model extraction technique.