Fast Machine Research Articles

Electrochemical reservoir computing based on surface-functionalized carbon nanotubes

Reservoir computing, a type of neural network computing, is an energy-efficient and fast machine learning method, expected to realize low-power edge processing in sensor devices. This study constructs an electrochemical reservoir using carbon nanotube (CNT) thin-film electrodes on flexible plastic and proves the capability of the electrochemical functionalization of CNT thin films to improve the reservoir's performance. The performance of the time-series data prediction task, called the nonlinear autoregressive moving average (NARMA) task, can be improved using multiple CNT electrodes functionalized under different conditions. Based on the correlation coefficient analysis of the reservoir output, we illustrate that the reservoir dimensionality is enhanced by combining CNT electrodes with different functionalization conditions. Furthermore, the dimensionality and memory capacity of the reservoir can be improved using the displacement current of the electric double layer and the redox current in the electrochemical electrode. We demonstrate the time-series data prediction of blood glucose levels in patients with type 1 diabetes using a physical reservoir for the first time. The use of functionalized CNTs enables flexible and high-performance reservoir computing, leading to the low-power in-sensor edge processing of wearable sensors.

A simple and reliable instance selection for fast training support vector machine: Valid Border Recognition

Support vector machines (SVMs) are powerful statistical learning tools, but their application to large datasets can cause time-consuming training complexity. To address this issue, various instance selection (IS) approaches have been proposed, which choose a small fraction of critical instances and screen out others before training. However, existing methods have not been able to balance accuracy and efficiency well. Some methods miss critical instances, while others use complicated selection schemes that require even more execution time than training with all original instances, thus violating the initial intention of IS. In this work, we present a newly developed IS method called Valid Border Recognition (VBR). VBR selects the closest heterogeneous neighbors as valid border instances and incorporates this process into the creation of a reduced Gaussian kernel matrix, thus minimizing the execution time. To improve reliability, we propose a strengthened version of VBR (SVBR). Based on VBR, SVBR gradually adds farther heterogeneous neighbors as complements until the Lagrange multipliers of already selected instances become stable. In numerical experiments, the effectiveness of our proposed methods is verified on benchmark and synthetic datasets in terms of accuracy, execution time and inference time.

Feature Weighting for Parkinson's Identification using Single Hidden Layer Neural Network

The diagnosis of Parkinson has become easier with the existence of machine learning. It includes using existing features from the biometric dataset generated by the person to identify whether he has Parkinson or not. The features differ in their discrimination capability and they suffer from redundancy. Hence, researchers have recommended using feature selection for Parkinson's identification. The feature selection aims at finding the most important and relevant features to produce an efficient and effective model. In this article, we present entropy-based Parkinson classification. The goal is to select only 50% of the most relevant features for Parkinson prediction. Two variants of neural networks are used for evaluation, the first one is a feed-forward Extreme Learning Machine ELM and the second one is Fast Learning Machine FLN. Also, the K-Nearest Neighbor KNN algorithm is used for evaluation. The results show the superiority of ELM and FLN when the model of feature selection is used with an accuracy of 80% compared with only 78% when the model is not used.

Pemodelan Prediksi Harga Ethereum (Atribut: Open, High dan Low) dengan Algoritma Extreme Learning Machine

The price of cryptocurrencies such as Ethereum often experiences high fluctuations and is difficult to predict. This study aims to predict Ethereum prices using the Extreme Learning Machine (ELM) algorithm which is a fast and efficient machine learning method. Ethereum price data is collected from CoinMarketCap by scraping the data using CoinmarketCap Scraper from the cryptocmd library using Python. An ELM model is built by changing the number of hidden nodes to determine the optimal prediction model of Ethereum prices based on the smallest average MAPE. Model performance was evaluated using the mean absolute percentage error (MAPE) on the test data set. The results show that the ELM model built can predict Ethereum prices with an accuracy of 96.96%. The MAPE obtained is 3.035334%, with 9 hidden nodes in the ELM network architecture model that was built. This shows that the model can explain about 96.96% of the variation in Ethereum price data. Therefore, the ELM model can be used as an aid in making investment decisions

Aspect-based classification of vaccine misinformation: a spatiotemporal analysis using Twitter chatter

BackgroundThe spread of misinformation of all types threatens people’s safety and interrupts resolutions. COVID-19 vaccination has been a widely discussed topic on social media platforms with numerous misleading and fallacious information. This false information has a critical impact on the safety of society as it prevents many people from taking the vaccine, decelerating the world’s ability to go back to normal. Therefore, it is vital to analyze the content shared on social media platforms, detect misinformation, identify aspects of misinformation, and efficiently represent related statistics to combat the spread of misleading information about the vaccine. This paper aims to support stakeholders in decision-making by providing solid and current insights into the spatiotemporal progression of the common misinformation aspects of the various available vaccines.MethodsApproximately 3800 tweets were annotated into four expert-verified aspects of vaccine misinformation obtained from reliable medical resources. Next, an Aspect-based Misinformation Analysis Framework was designed using the Light Gradient Boosting Machine (LightGBM) model, which is one of the most advanced, fast, and efficient machine learning models to date. Based on this dataset, spatiotemporal statistical analysis was performed to infer insights into the progression of aspects of vaccine misinformation among the public. Finally, the Pearson correlation coefficient and p-values are calculated for the global misinformation count against the vaccination counts of 43 countries from December 2020 until July 2021.ResultsThe optimized classification per class (i.e., per an aspect of misinformation) accuracy was 87.4%, 92.7%, 80.1%, and 82.5% for the “Vaccine Constituent,” “Adverse Effects,” “Agenda,” “Efficacy and Clinical Trials” aspects, respectively. The model achieved an Area Under the ROC Curve (AUC) of 90.3% and 89.6% for validation and testing, respectively, which indicates the reliability of the proposed framework in detecting aspects of vaccine misinformation on Twitter. The correlation analysis shows that 37% of the countries addressed in this study were negatively affected by the spread of misinformation on Twitter resulting in reduced number of administered vaccines during the same timeframe.ConclusionsTwitter is a rich source of insight on the progression of vaccine misinformation among the public. Machine Learning models, such as LightGBM, are efficient for multi-class classification and proved reliable in classifying vaccine misinformation aspects even with limited samples in social media datasets.

Study on the Performance of laser-assisted fast tool servo machining device for glass-ceramic

As an optical hard brittle material, glass-ceramic is widely used in radome, astronomical telescopes, and laser industries. However, an efficient method for machining optical free-form surface of glass-ceramic still needs to be studied by academia and industry. In this study, a laser-assisted fast tool servo machining device for glass-ceramic is developed. The theoretical stiffness of semicircular flexure hinge in the device is calculated by flexibility matrix method. The displacement output characteristics of the device are verified by simulation and displacement measurement experiment. Static and dynamic temperature measurement experiments of glass-ceramic are carried out respectively. The free-form surface morphology of glass-ceramic is obtained through machining experiment. Laser-assisted fast tool servo machining can reduce surface roughness by 24.4% and tool wear by 35.7% compared to conventional machining, which fully proves that the laser-assisted fast tool servo machining device can obtain higher machining quality of optical free-form surface.

Compressed computations using wavelets for hidden Markov models with continuous observations.

Compression as an accelerant of computation is increasingly recognized as an important component in engineering fast real-world machine learning methods for big data; c.f., its impact on genome-scale approximate string matching. Previous work showed that compression can accelerate algorithms for Hidden Markov Models (HMM) with discrete observations, both for the classical frequentist HMM algorithms-Forward Filtering, Backward Smoothing and Viterbi-and Gibbs sampling for Bayesian HMM. For Bayesian HMM with continuous-valued observations, compression was shown to greatly accelerate computations for specific types of data. For instance, data from large-scale experiments interrogating structural genetic variation can be assumed to be piece-wise constant with noise, or, equivalently, data generated by HMM with dominant self-transition probabilities. Here we extend the compressive computation approach to the classical frequentist HMM algorithms on continuous-valued observations, providing the first compressive approach for this problem. In a large-scale simulation study, we demonstrate empirically that in many settings compressed HMM algorithms very clearly outperform the classical algorithms with no, or only an insignificant effect, on the computed probabilities and infered state paths of maximal likelihood. This provides an efficient approach to big data computations with HMM. An open-source implementation of the method is available from https://github.com/lucabello/wavelet-hmms.

Fast and accurate machine learning prediction of phonon scattering rates and lattice thermal conductivity

Lattice thermal conductivity is important for many applications, but experimental measurements or first principles calculations including three-phonon and four-phonon scattering are expensive or even unaffordable. Machine learning approaches that can achieve similar accuracy have been a long-standing open question. Despite recent progress, machine learning models using structural information as descriptors fall short of experimental or first principles accuracy. This study presents a machine learning approach that predicts phonon scattering rates and thermal conductivity with experimental and first principles accuracy. The success of our approach is enabled by mitigating computational challenges associated with the high skewness of phonon scattering rates and their complex contributions to the total thermal resistance. Transfer learning between different orders of phonon scattering can further improve the model performance. Our surrogates offer up to two orders of magnitude acceleration compared to first principles calculations and would enable large-scale thermal transport informatics.

Study of the surface roughness and optimization of machining parameters during laser-assisted fast tool servo machining of glass-ceramic

Glass-ceramic is a typical hard and brittle material that is difficult to machine. In order to improve the surface quality of laser-assisted fast tool servo machining optical free-form surface of glass-ceramic, the effects of spindle speed, feed speed, piezoelectric frequency and laser power on the surface roughness were investigated. Firstly, the Taguchi method (TM) was used to establish the orthogonal experiment, and the contribution rate of each machining parameter to the surface roughness was obtained through variance and signal-to-noise ratio (S/N) analysis. The order of the influence degree of each parameter on the surface roughness is as follows: laser power > spindle speed > feed speed > piezoelectric frequency. The optimal machining parameter combinations obtained for the TM experiment are as follows: spindle speed 50 rpm, feed speed 0.01 mm rev−1, piezoelectric frequency 8 Hz, laser power 75 W. The range of surface roughness reduction obtained by comparing laser-assisted machining (LAM) with pure fast tool servo (FTS) machining is 38.75%∼58.77%. The Box-Behnken Design (BBD) in response surface methodology (RSM) was used to design experiments and a regression model for surface roughness was established through RSM. The deviation between the surface roughness predicted by the regression equation and the experimental value is less than ±6%. The influence law of various machining parameters on surface roughness was studied through three-dimensional response surface. RSM optimized the minimum surface roughness with a desirability of 99.43%. The optimal combination of machining parameters optimized through RSM is as follows: spindle speed 53.71 rpm, feed speed 0.02 mm rev−1, piezoelectric frequency 6.73 Hz, laser power 72 W. This paper is the first to combine LAM with FTS for machining optical free-form surface of glass-ceramic. This study provides a reference for laser-assisted fast tool servo machining and the research methods of surface quality.

Suggestion of new models for predicting performance of raise boring machines based on indentation tests

Raise Boring Machines (RBMs) as the fast, safe, and efficient machines are used for drilling/excavation of vertical and/or inclined shafts in the underground structures. Proper selection and accurate performance prediction of these machines are very important issues affecting the cost estimation and planning of the projects. However, due to lack of reliable models, performance prediction of these machines has been seriously limited. This study aims at proposing new models for performance prediction of RBMs based on indentation and full-scale linear cutting tests, as well as physical–mechanical property tests, performed on rock samples. Operational and performance data of RBMs and rock samples with different geological origins are collected from different raise boring projects. Then, indentation tests are performed on different core samples by using a real-life tungsten carbide insert of the button cutters used on the analyzed RBMs. Full-scale linear cutting tests were previously performed on the block samples by using a 305 mm (12 in.) diameter button (kerf) cutter. The study indicates that velocity of S-wave and Cerchar abrasivity index can be used to estimate brittleness index of rocks, which is also obtained from the indentation tests. It also shows the feasibility of using power function as relationship between the indentation force and penetration values for performance prediction of RBMs. It is shown that the results of indentation tests can be used in both empirical and deterministic predictions of RBM performance. However, the proposed models should be improved by additional rock types, operational-performance parameters of different RBMs, indentation and cutting test results of different rocks.

PERMANENT MAGNET SYNCHRONOUS MOTOR TORQUE RIPPLE REDUCTION USING PREDICTIVE TORQUE CONTROL

The control of (PMSM) is the subject of this study and the torque ripple reduction in (PMSM) is the main goal of this work. Torque and flux are controlled using a predictive model and Vector Control. Because it is commonly employed in regulating electric motors. Space vector control and predictive control are two (PMSM) control approaches used in this study. Predictive control was determined to be more effective in terms of response and action after a Matlab simulation of the two approaches. MPC covers all potential switching states that decrease actual torque and flux ripples as well as Total Harmonic Distortion. The benefits of MPC include simple principles, an easy-to-use console, and the ability to implement limits quickly. However, there are some drawbacks to this technique, including the requirement for bigger accounts and faster machines. The fundamental principles of the control techniques discussed are provided A (PMSM) powered by a two-level power converter is then used to simulate the control approaches. Their performance in comparison is based on the obtained results.

Recent Roles of Artificial Intelligence Artists in Art Circulation

In this modern world, while people are used to living a life of competition and a fast environment, artificial intelligence (AI) and machine learning have been established to serve people. AI and machine learning developed on the concept of repetition and learning. Now some companies use AI to replace or support humans in jobs that operate according to patterns. AI and machine learning are unlike humans, they have no bias since they are not exhausted. Also, employers do not have to pay for overtime. Some companies even have AI to calculate risk and return for management to make decisions. In recent years, many companies start considering using AI and machine learning to replace human employees for positions that do not require particular special skills which need by humans. Recently, machine learning has been developed to generate pictures from the text. This means even people who are not good at art at all can command AI to generate pictures, particularly for them. Accessing those AI is simply just by logging in to the program online. When generating pictures is so easy as that, what will AI artists play next role in art circulation? In this paper, we conduct mixed methodology by qualitative and quantitative methods to research the tendency of AI artists in the future.

Fast One-Sided RDMA-Based State Machine Replication for Disaggregated Memory

Disaggregated memory architecture has risen in popularity for large datacenters with the advantage of improved resource utilization, failure isolation, and elasticity. Replicated state machines (RSMs) have been extensively used for reliability and consistency. In traditional RSM protocols, each replica stores replicated data and has the computing power to participate in some part of the protocols. However, traditional RSM protocols fail to work in the disaggregated memory architecture due to asymmetric resources on CPU nodes and memory nodes. This article proposes ECHO, a fast one-sided RDMA-based RSM protocol with lightweight log replication and remote applying, efficient linearizability guarantee, and fast coordinator failure recovery. ECHO enables all operations in the protocol to be efficiently executed using only one-sided RDMA, without the participation of any computing resource in the memory pool. To provide lightweight log replication and remote applying, ECHO couples the replicated log and the state machine to avoid dual-copy and performs remote applying by updating pointers. To enable efficient remote log state management, ECHO leverages a hitchhiked log state updating scheme to eliminate extra network round trips. To provide efficient linearizability guarantee, ECHO performs immediate remote applying after log replication and leverages the local locks at the coordinator to ensure linear consistency. Moreover, ECHO adopts a commit-aware log cache to make data visible immediately after being committed. To achieve fast failure recovery, ECHO leverages a commit point identification scheme to reduce the overhead of log consistency recovery. Experimental results demonstrate that ECHO outperforms the state-of-the-art RSM protocol (namely Sift) in multiple scenarios. For example, ECHO achieves 27%–52% higher throughput on typical write-intensive workloads. Moreover, ECHO reduces the consistency recovery time by three orders of magnitude for coordinator failure.

CoolPINNs: A physics-informed neural network modeling of active cooling in vascular systems

Emerging technologies like hypersonic aircraft, space exploration vehicles, and batteries avail fluid circulation in embedded microvasculatures for efficient thermal regulation. Modeling is vital during the design and operational phases of these engineered systems. However, many challenges exist in developing a modeling framework. What is lacking is an accurate framework that (i) captures sharp jumps in the thermal flux across complex vasculature layouts, (ii) deals with oblique derivatives (involving tangential and normal components), (iii) handles nonlinearity because of radiative heat transfer, (iv) provides a high-speed forecast for real-time monitoring, and (v) facilitates robust inverse modeling. This paper addresses these challenges by availing the power of physics-informed neural networks (PINNs). We develop a fast, reliable, and accurate Scientific Machine Learning (SciML) framework for vascular-based thermal regulation—called CoolPINNs: a PINNs-based modeling framework for active cooling. The proposed mesh-less framework elegantly overcomes all the mentioned challenges. The significance of the reported research is multi-fold. First, the framework is valuable for real-time monitoring of thermal regulatory systems because of rapid forecasting. Second, researchers can address complex thermoregulation designs since the approach is meshless. Finally, the framework facilitates systematic parameter identification and inverse modeling studies, perhaps the most significant utility of the current framework.

Broad Graph Convolutional Neural Network and Its Application in Hyperspectral Image Classification

A fast and effective machine learning method, the broad learning system (BLS), has been successfully used for hyperspectral image (HSI) classification with good results. However, the original BLS cannot fully utilize the spatial information of HSI, and the linear sparse features of mapping nodes (MFs) have insufficient ability to characterize HSI. Thus, a broad graph convolutional neural network (BGCNN) is proposed for solving the aforementioned issues. In the BGCNN, the graph convolution operation is first used to capture the nonlinear spectral-spatial features, instead of only the linear sparse autoencoder in BLS. Then, the spectral-spatial features are expanded with a graph convolution operation, which further enhances the feature representation capability. Finally, the ridge regression theory is exploited to acquire the output weights. Experiments on four real HSI datasets show that our proposed BGCNN outperforms several state-of-the-art classification methods on the classification accuracy with a relatively less consumed time.

Machine learning and reduced order modelling for the simulation of braided stent deployment.

Endoluminal reconstruction using flow diverters represents a novel paradigm for the minimally invasive treatment of intracranial aneurysms. The configuration assumed by these very dense braided stents once deployed within the parent vessel is not easily predictable and medical volumetric images alone may be insufficient to plan the treatment satisfactorily. Therefore, here we propose a fast and accurate machine learning and reduced order modelling framework, based on finite element simulations, to assist practitioners in the planning and interventional stages. It consists of a first classification step to determine a priori whether a simulation will be successful (good conformity between stent and vessel) or not from a clinical perspective, followed by a regression step that provides an approximated solution of the deployed stent configuration. The latter is achieved using a non-intrusive reduced order modelling scheme that combines the proper orthogonal decomposition algorithm and Gaussian process regression. The workflow was validated on an idealized intracranial artery with a saccular aneurysm and the effect of six geometrical and surgical parameters on the outcome of stent deployment was studied. We trained six machine learning models on a dataset of varying size and obtained classifiers with up to 95% accuracy in predicting the deployment outcome. The support vector machine model outperformed the others when considering a small dataset of 50 training cases, with an accuracy of 93% and a specificity of 97%. On the other hand, real-time predictions of the stent deployed configuration were achieved with an average validation error between predicted and high-fidelity results never greater than the spatial resolution of 3D rotational angiography, the imaging technique with the best spatial resolution (0.15mm). Such accurate predictions can be reached even with a small database of 47 simulations: by increasing the training simulations to 147, the average prediction error is reduced to 0.07mm. These results are promising as they demonstrate the ability of these techniques to achieve simulations within a few milliseconds while retaining the mechanical realism and predictability of the stent deployed configuration.

An intelligent algorithm for fast machine translation of long English sentences

Abstract Translation of long sentences in English is a complex problem in machine translation. This work briefly introduced the basic framework of intelligent machine translation algorithm and improved the long short-term memory (LSTM)-based intelligent machine translation algorithm by introducing the long sentence segmentation module and reordering module. Simulation experiments were conducted using the public corpus and the local corpus containing self-collected linguistic data. The improved algorithm was compared with machine translation algorithms based on a recurrent neural network and LSTM. The results suggested that the LSTM-based machine translation algorithm added with the long sentence segmentation module and reordering module effectively segmented long sentences and translated long English sentences more accurately, and the translation was more grammatically correct.

Compensation machining of freeform based on fast tool servo system.

With the advantages of large field of view, low cost and simple structure, the freeform optical system has extensive requirements in space exploration and other fields. However, the current machining methods for freeform are difficult to meet the requirements of optical use. Based on a developed fast servo tool (FTS) device, this paper proposes an error compensation turning method for freeforms. Firstly, the Zernike polynomial fitting method is used to reconstruct the freeform surface shape error obtained from off-line measurement, and the offset compensation is used to correct the tool path. Then, the compensation processing physical system is built to simulate the off-line compensation processing of the workpiece to verify the feasibility of compensation processing. Finally, the turning compensation processing of convex freeform aluminum mirror is carried out, and the surface accuracy of the workpiece meets the requirements of visible band. The research results have important practical significance for realizing the fast response machining of free-form surface mirror.

Fast support vector machine training via three-term conjugate-like SMO algorithm

Support vector machine (SVM) is an important class of methods in pattern recognition, and the sequential minimal optimization (SMO) algorithm is one of the most popular methods for training SVM at present. Based on the conjugate sequential minimal optimization algorithm (CSMO), we propose a novel three-term conjugate-like sequential minimal optimization algorithm (TCSMO) for classification and regression tasks. Compared with the CSMO, although the three-term conjugate-like SMO slightly increases the amount of arithmetic operations in each iteration, it significantly reduces the number of iterations required to converge to the specified accuracy and shortens the training time of the SVM. Additionally, we give a convergence proof of the three-term conjugate-like SMO algorithm and four new conjugate parameters. Numerical experiments show that the three-term conjugate-like SMO algorithm performs better numerically in both classification and regression tasks.

Nanosecond Jet Classification at LHC

We investigate the possibility of using Deep Learning algorithms for jet identification in the L1 trigger at HL-LHC. We perform a survey of architectures (MLP, CNN, Graph Networks) and benchmark their performance and resource consumption on FPGAs using a QKeras for quantization aware training and HLS4ML to convert the neural network model into a FPGA firmware. We use the HLS4ML jet dataset to compare the results obtained in this study to previous literature on Fast Machine Learning applications on FPGAs.