FPGA-based implementation of classification techniques: A survey

Recently, a number of classification techniques have been introduced. However, processing large dataset in a reasonable time has become a major challenge. This made classification task more complex and expensive in calculation. Thus, the need for solutions to overcome these constraints such as field programmable gate arrays (FPGAs). In this paper, we give an overview of the various classification techniques. Then, we present the existing FPGA based implementation of these classification methods. After that, we investigate the confronted challenges and the optimizations strategies. Finally, we highlight the hardware accelerator architectures and tools for hardware design suggested to improve the FPGA implementation of classification methods.     

Matrix Manipulation of Directly-Synthesized PbS Quantum Dot Inks Enabled by Coordination Engineering

The direct-synthesis of conductive PbS quantum dot (QD) ink is facile, scalable, and low-cost, boosting the future commercialization of optoelectronics based on colloidal QDs. However, manipulating the QD matrix structures still is a challenge, which limits the corresponding QD solar cell performance. Here, for the first time a coordination-engineering strategy to finely adjust the matrix thickness around the QDs is presented, in which halogen salts are introduced into the reaction to convert the excessive insulating lead iodide into soluble iodoplumbate species. As a result, the obtained QD film exhibits shrunk insulating shells, leading to higher charge carrier transport and superior surface passivation compared to the control devices. A significantly improved power-conversion efficiency from 10.52% to 12.12% can be achieved after the matrix engineering. Therefore, the work shows high significance in promoting the practical application of directly synthesized PbS QD inks in large-area low-cost optoelectronic devices.     

Temperature dependent negative permittivity in solid solutions Sr2Mn1-xSnxO4 (x = 0, 0.3, 0.5)

A few compositions of the system Sr₂Mn_1-xSn_xO₄ (x = 0.0, 0.3, 0.5) were synthesized in the air by the solid-state ceramic route. A change in the sign (positive to negative) of the permittivity above a particular temperature (T_C) is observed at all the measured frequencies. The negative permittivity was analyzed by the Drude-Lorentz model. It was found that negative permittivity is caused by the plasma oscillations of thermally excited free charge carriers. Analysis of XPS spectra confirmed the presence of mixed-valence states of both Mn (Mn⁴⁺ and Mn³⁺) and Sn (Sn⁴⁺ and Sn²⁺) ions. The UV–vis.-IR spectroscopy results indicated generation of a large number of defect states in the forbidden bandgap region of Sr₂MnO₄ on the substitution of Sn at Mn site. Synthesized samples are promising metamaterials for radio frequency (10 Hz -2 MHz) region applications due to the high-temperature plasmonic behavior.     

Grain orientation evolution and multi-scale interfaces enhanced thermoelectric properties of textured Sr0.9La0.1TiO3 based ceramics

Sr_0.9La_0.1TiO₃ based textured ceramics (SLTT-S3T) with a texture fraction of 0.81 are successfully fabricated by the reactive template grain growth method, in which Sr_0.9La_0.1TiO₃/20 wt%Ti was used as matrix and 10 wt% plate-like Sr₃Ti₂O₇ template seeds were used as templates. The phase transition, microstructure evolution, and the anisotropic thermoelectric properties of SLTT-S3T ceramics were investigated. The results show that the ceramics are mainly composed of Sr_0.9La_0.1TiO₃ and rutile TiO₂ phases. Grains grow with a preferred orientation along (h00). A maximum ZT of 0.26 at 1073 K was achieved in the direction perpendicular to the tape casting direction. The low lattice thermal conductivity of 1.9 W/(m K) at 1073 K was obtained decreased by 34%, 40%, and 38% compared with non-textured, SrTiO₃ and Sr_0.9La_0.1TiO₃ ceramics prepared by the same process, can be attributed to the enhanced phonon scattering by the complex multi-scale boundaries and interfaces. This work provides a strategy of microstructural design for thermoelectric oxides to decrease intrinsic lattice thermal conductivity and further regulate thermoelectric properties via texture engineering.     

Effects of hydrogen plasma treatment on the physical and chemical properties of tin oxide thin films for ambipolar thin-film transistor applications

In this study, we investigated the physical and chemical properties of H₂ plasma-treated tin oxide (SnO_X) thin films, followed by their applications in ambipolar thin-film transistors (TFTs). Finely controlled H₂ implantation was carried out using a reactive-ion-etching system at a radio frequency power of 30 W and under various exposure times. H₂ plasma treatments induced changes in the chemical structures and surface morphologies of the SnO_X thin films, including a partial phase transformation of Sn and SnO to SnO₂. The defects originating from oxygen vacancies (O_Vacs) in the SnO_X thin films were passivated by H via the formation of Sn–H bonds, which decreased the density of subgap states in the SnO_X thin films. The H₂ plasma-treated SnO_X TFTs showed considerably improved ambipolarity and electrical performance. Complementary metal–oxide–semiconductor (CMOS) logic inverters comprising H₂-plasma-treated ambipolar SnO_X TFTs exhibited a maximum gain of 34.5 V/V at a supply voltage of 10 V. The results of this study present the meaningful investigation of H₂ plasma-treated ambipolar SnO_X TFTs that can be used to fabricate CMOS circuits for various applications.     

Improving accuracy of automatic optical inspection with machine learning

Electronic devices require the printed circuit board(PCB)to support the whole structure,but the assembly of PCBs suffers from welding problem of the electronic components such as surface mounted devices(SMDs)resistors.The automated optical inspection(AOI)machine,widely used in industrial production,can take the image of PCBs and examine the welding issue.However,the AOI machine could commit false negative errors and dedicated technicians have to be employed to pick out those misjudged PCBs.This paper proposes a machine learning based method to improve the accuracy of AOI.In particular,we propose an adjacent pixel RGB value based method to pre-process the image from the AOI machine and build a customized deep learning model to classify the image.We present a practical scheme including two machine learning procedures to mitigate AOI errors.We conduct experiments with the real dataset from a production line for three months,the experimental results show that our method can reduce the rate of misjudgment from 0.3%–0.5%to 0.02%–0.03%,which is meaningful for thousands of PCBs each containing thousands of electronic components in practice.     

Fabrication of high-efficiency Yb:Y2O3 laser ceramics without photodarkening

High-efficiency Yb:Y₂O₃ laser ceramics were fabricated using the vacuum-sintering plus hot isostatic pressing (HIP) without sintering additives. High-purity well-dispersed nanocrystalline Yb:Y₂O₃ powder was synthesized using a modified co-precipitation method in-house. The green bodies were first vacuum sintered at a temperature as low as 1430°C and then HIPed at 1450°C. Finally, the samples were air annealed at 800°C for 10 h. Although no sintering aids were used, full density of the samples with excellent optical homogeneity and an inline transmission of 80% at 400 nm could be obtained. Moreover, photodarkening phenomenon was not detected in the ceramics. Preliminary laser experiment with the fabricated ceramics in a two-mirror cavity has demonstrated 32 W continuous-wave (CW) output at ∼1077 nm with an optical-to-optical conversion efficiency of 58.2%. To the best of our knowledge, this is so far the highest CW output power and optical-to-optical conversion efficiency achieved with the Yb³⁺-doped sesquioxide ceramics in a simple two-mirror cavity.     

Finite-time continuous extended state observers: design and experimental validation on electro-hydraulic systems✰

In this paper, a time-varying chattering-free observer for electro-hydraulic actuators able to provide finite-time estimations of the full state variables as well as uncertainties is presented. The key idea is to employ a positive-increasing function associated with the observer objectives to improve the convergence time performance. First, the model of an electro-hydraulic actuator as a case study for the proposed observer is presented. Then, an extended state observer is introduced, and the time-varying gains are provided to ensure that without neither any knowledge about the upper bounds of the uncertainties nor its derivative, the observation error dynamics is convergent to a neighborhood of zero in a finite time. Comparative simulations are presented to analyze the effectiveness of the proposed observers. Then, further simulations are performed in the presence of measurement noise. It is concluded that the proposed scheme can compete with other leading strategies, making it a qualified alternative approach in the observer design with noteworthy potential. Finally, the effectiveness of the proposed approach in real-life conditions is demonstrated through experimental studies.     

Mutual information maximizing GAN inversion for real face with identity preservation

Recent generative adversarial networks (GANs) have yielded remarkable performance in face image synthesis. GAN inversion embeds an image into the latent space of a pretrained generator, enabling it to be used for real face manipulation. However, current inversion approaches for real faces suffer the dilemma of initialization collapse and identity loss. In this paper, we propose a hierarchical GAN inversion for real faces with identity preservation based on mutual information maximization. We first use a facial domain guaranteed initialization to avoid the initialization collapse. Furthermore, we prove that maximizing the mutual information between inverted faces and their identities is equivalent to minimizing the distance between identity features from inverted and original faces. Optimization for real face inversion with identity preservation is implemented on this mutual information-maximizing constraint. Extensive experimental results show that our approach outperforms state-of-the-art solutions for inverting and editing real faces, particularly in terms of face identity preservation.