Feedforward Model Research Articles

Prediction of Women’s Fashion Product Expense in Hanoi City Using Artificial Neural Network

A person's fashion product expense plays an important role in building a garment company's product and pricing strategy. This article presents the results of predicting the fashion expenses of women aged 25 to 55 in Hanoi city. A linear multivariable model determined by the Bayesion Model Average (BMA) was applied to predict expenditures based on a set of 150 women survey data. The 3 layers feed-forward neural network model with 8 input neurons, 4 (+1) hidden layer neurons, and 1 output layer neuron, The Sigmoid activation function, trained by the error backpropagation algorithm has been established on software R specifically predicts the value of women's fashion expense with the above data set. The results show that the artificial neural network (ANN) has been set up allowing to prediction of fashion expenses accurately. The expenditure values predicted by the ANN and the actual value are linearly correlated with R2 = 0.987; The prediction results by ANN are more accurate than the defined multivariable linear model.

Topology Optimization of Wing Ribs for Additive Manufacturing

This paper describes the design of a lightweight wing rib structure by combining topology optimisation with additive manufacturing. In addition, a deep feed-forward neural network model is proposed to perform the load prediction for the constructed wing structure incorporated with this optimised rib. The strain energy of the front rib has a minimum strain energy 1330 J for the initial volume state when the topology is not optimised for lightweighting. The relative error of the load prediction values obtained by the output layer of the deep feed-forward neural network is less than 0.02%. The absolute error of the small load prediction was less than 0.30 N. The presented results demonstrate the viability of additive manufactured rib and its implementation in global wing load prediction model for faster designs.

Optimization of neural-network model using a meta-heuristic algorithm for the estimation of dynamic Poisson’s ratio of selected rock types

This research focuses on the predictive modeling between rocks' dynamic properties and the optimization of neural network models. For this purpose, the rocks' dynamic properties were measured in terms of quality factor (Q), resonance frequency (FR), acoustic impedance (Z), oscillation decay factor (α), and dynamic Poisson’s ratio (v). Rock samples were tested in both longitudinal and torsion modes. Their ratios were taken to reduce data variability and make them dimensionless for analysis. Results showed that with the increase in excitation frequencies, the stiffness of the rocks got increased because of the plastic deformation of pre-existing cracks and then started to decrease due to the development of new microcracks. After the evaluation of the rocks’ dynamic behavior, the v was estimated by the prediction modeling. Overall, 15 models were developed by using the backpropagation neural network algorithms including feed-forward, cascade-forward, and Elman. Among all models, the feed-forward model with 40 neurons was considered as best one due to its comparatively good performance in the learning and validation phases. The value of the coefficient of determination (R2 = 0.797) for the feed-forward model was found higher than the rest of the models. To further improve its quality, the model was optimized using the meta-heuristic algorithm (i.e. particle swarm optimizer). The optimizer ameliorated its R2 values from 0.797 to 0.954. The outcomes of this study exhibit the effective utilization of a meta-heuristic algorithm to improve model quality that can be used as a reference to solve several problems regarding data modeling, pattern recognition, data classification, etc.

University Auto Reply FAQ Chatbot Using NLP and Neural Networks

When new students enter college, they often have similar questions – “Where to study for this subject?,” “How to prepare Data Structures and Algorithms?,” “How to connect with seniors?,” and so on. The use of chatbots can help them get answers to their questions quickly and efficiently. This study proposes a deep learning (DL) chatbot for addressing common doubts of university students, providing efficient and accurate responses to college-specific questions. A self-curated dataset is used for the purpose of building the chatbot, and natural language processing techniques are utilized for the pre-processing of raw data gathered. The study compares two deep learning models – a bidirectional long- and short-term memory network and a simple feed-forward neural network model.

التنبُّؤ بسعر الصرف الفوري لليورو دولار في سوق (FOREX) استعمال المفاضلة بين طريقة هولت والشبكات العصبية الاصطناعية

In this paper , the time series data of the daily exchange rate of the euro against the dollar were used for the period from (12/7/2020 H:1 AM-10/12/2020 H:7 AM) with a total of (79) Observations as (120) Observations for comparison with the forecast values obtained using the two methods, to make the comparison process for future forecasts for the period from (10/12/2020 H:8 AM- 15/12/2020 H:7 AM) with a total of (120) Observations. To predict them, between the double exponential smoothing model (Holt method) and the feed-forward artificial neural network model using the two algorithms (Incremental Back Propagation algorithm, Quick Propagation algorithm).These methods are characterized by high accuracy and flexibility of these methods in the process of analyzing the time series, the results of the application showed that the most efficient and optimal model for representing the time series data is the artificial neural network model [2-10-1] using the Quick Propagation algorithm for the daily exchange rate of the euro against the dollar according to the criterion The mean square error ( MSE), has given lower indicators than the indicators of the artificial neural network model [2-10-1] using the Incremental Back Propagation algorithm, and the double exponential smoothing model (Holt method) when using (α =0.9 and = 0.1), which clearly indicates that it is the appropriate and efficient model for estimating future forecasts for the period from (10/12/2020 H:8 AM- 15/12/2020 H:7 AM). Where these values showed consistency with their counterparts in the original series, and provided us with a future picture of the reality of the daily exchange rate of the euro against the dollar for that period. Therefore, the artificial neural network model provided better future predictions than those provided by the double exponential smoothing model (Holt method), according to the standard of mean square error ( MSE), it gave less indicators than the indicators of the Holt model. The ready-made statistical programs MinitabV18 were used in the statistical side, and the ready-made neural network system program known as Alyuda NeuroIntelligence was used in the neural networks side.

Predictive Control of a Wind Turbine Based on Neural Network-Based Wind Speed Estimation

Predictive control is an advanced control technique that performs well in various application domains. In this work, linearised control design models are first derived in state-space form from the full nonlinear model of the 5 MW Supergen (Sustainable Power Generation and Supply) exemplar wind turbine. Feedback model predictive controllers (FB-MPCs) and feedforward model predictive controllers (FF-MPCs) are subsequently designed based on these linearised models. A neural network (NN)-based wind speed estimation method is then employed to obtain the accurate wind estimation required for designing a FF-MPC. This method uses a LiDAR to be shared between multiple wind turbines in a cluster, i.e., one turbine is mounted with a LiDAR, and each of the remaining turbines from the cluster is provided with a NN-based estimator, which replaces the LiDAR, making the approach more economically viable. The resulting controllers are tested by application to the full nonlinear model (based on which the linearised models are derived). Moreover, the mismatch between the control design model and the simulation model (model–plant mismatch) allows the robustness of the controllers’ design to be tested. Simulations are carried out at varying wind speeds to evaluate the robustness of the controllers by applying them to a full nonlinear 5 MW Matlab/SIMULINK model of the same exemplar Supergen wind turbine. Improved torque/speed plane tracking is achieved with a FF-MPC compared to a FB-MPC. Simulation results further demonstrate that the control performance is enhanced in both the time and frequency domains without increasing the wind turbine’s control activity; that is, the controller’s gain crossover frequency (or bandwidth) remains within the acceptable range, which is about 1 rad/s.

Maternal anxiety-driven modulation of fetal limbic connectivity designs a backbone linking neonatal brain functional topology to socio-emotional development in early childhood.

A link between maternal anxiety during pregnancy and adverse socio-emotional outcomes in childhood has been consistently sustained on the very early neurodevelopmental alteration of structural pathways between fetal limbic and cortical brain regions. In this study, we provide follow-up evidence for a feed-forward model linking (i) maternal anxiety, (ii) fetal functional neurodevelopment, (iii) neonatal functional network organization with (iv) socio-emotional neurobehavioral development in early childhood. Namely, we investigate a sample of 16 mother-fetus dyads and show how a maternal state-trait anxiety profile with pregnancy-specific worries can significantly influence functional synchronization patterns between regions of the fetal limbic system (i.e., hippocampus and amygdala) and the neocortex, as assessed through resting-state functional magnetic resonance imaging. Generalization of the findings was supported by leave-one-out cross-validation. We further show how this maternal-fetal cross-talk propagates to functional network topology in the neonate, specifically targeting connector hubs, and further maps onto socio-emotional profiles, assessed through Bayley-III socio-emotional scale in early childhood (i.e., in the 12-24 months range). Based on this evidence, we put forward the hypothesis of a "Maternal-Fetal-Neonatal Anxiety Backbone", through which neurobiological changes driven by maternal anxiety could trigger a divergence in the establishment of a cognitive-emotional development blueprint, in terms of the nascent functional homeostasis between bottom-up limbic and top-down higher-order neuronal circuitry.

A simple ANN-MLP model for estimating 60-GHz PDP inside public and private vehicles

Radio wave propagation in an intra-vehicular (IV) environment is markedly different from other well-studied indoor scenarios, such as an office or a factory floor. While millimetre wave (mmWave)-based intra-vehicular communications promise large bandwidth and can achieve ultra-high data rates with lower latency, exploiting the advantages of mmWave communications largely relies on adequately characterising the propagation channel. Channel characterisation is most accurately done through an extensive channel sounding, but due to hardware and environmental constraints, it is impractical to test channel conditions for all possible transmitter and receiver locations. Artificial neural network (ANN)-based channel sounding can overcome this impediment by learning and estimating the channel parameters from the channel environment. We estimate the power delay profile in intra-vehicular public and private vehicle scenarios with a high accuracy using a simple feedforward multi-layer perception-based ANN model. Such artificially generated models can help extrapolate other relevant scenarios for which measurement data are unavailable. The proposed model efficiently matches the taped delay line samples obtained from real-world data, as shown by goodness-of-fit parameters and confusion matrices.

Stochastic modeling of physical drag coefficient – Its impact on orbit prediction and space traffic management

Ambitious satellite constellation projects by commercial entities and the ease of access to space in recent times have led to a dramatic proliferation of low-Earth space traffic. It jeopardizes space safety and long-term sustainability, necessitating better space domain awareness (SDA). Correct modeling of uncertainties in force models and orbital states, among other things, is an essential part of SDA. For objects in the low-Earth orbit (LEO) region, the uncertainty in the orbital dynamics mainly emanate from limited knowledge of the atmospheric drag-related parameters and variables. In this paper, which extends the work by Paul et al. (2021), we develop a feed-forward deep neural network model for the prediction of the satellite drag coefficient for the full range of satellite attitude (i.e., satellite pitch ∈ (-90°,+90°) and satellite yaw ∈ (0°,+360°)). The model simultaneously predicts the mean and the standard deviation and is well-calibrated. We use numerically simulated physical drag coefficient data for training our neural network. The numerical simulations are carried out using the test particle Monte Carlo method using the diffuse reflection with incomplete accommodation gas-surface interaction model. Modeling is carried out for the well-known CHAllenging Minisatellite Payload (CHAMP) satellite. Finally, we use the Monte Carlo approach to propagate CHAMP over a three-day period under various modeling scenarios to investigate the distribution of radial, along-track, and cross-track orbital errors caused by drag coefficient uncertainty. The key takeaways of this paper are - (a) a constant drag coefficient cannot be used for reliable SDA purposes, and (b) stochastic machine learning models allow for the computation of drag coefficients in a timely manner while providing reliable uncertainty estimates.

Integrating experiments, finite element analysis, and interpretable machine learning to evaluate the auxetic response of 3D printed re-entrant metamaterials

Metamaterials have received extensive attention in fundamental and applied research over the past two decades due to their unique mechanical behavior. This paper presents an interpretable machine learning (ML) approach for efficient response prediction of three-dimensional (3D)-printed metamaterials. However, developing such an ML-based model requires a large consistent, representative, balanced, and complete dataset. To this extent, an experimentally validated finite element analysis (FEA) approach is implemented to generate 8096 non-self-intersecting re-entrant honeycomb structures by varying the mesoscale geometrical features to obtain the corresponding Poisson's ratios. This dataset is leveraged to develop a feed-forward multilayer perceptron-based predictive model. The developed ML model shows excellent predictive efficacy on the unseen test dataset. Shapely additive explanation (SHAP) is then used for model interpretation. SHAP results show that the slant cell length is the dominant input feature dictating the model output whereas cell angle and vertical cell length show mixed trends signifying that other input features influence their effect on the model output. Moreover, cell thickness does not significantly influence the model output when compared to other input features. Overall, the integrated numerical simulation-experiment-interpretable ML-based predictive approach presented here can be leveraged to design and develop metamaterials for a wide range of engineering applications.

Target-learning the Latent Space of a Variational Autoencoder model for the Inverse Design of Stable Perovskites

At the forefront of discoverable materials are perovskites that stand out as some of the most chemically diverse and multifunctional energy materials. Theoretically, the estimated number of ternary perovskites exceeds a hundred thousand distinct compounds, notwithstanding that only a small fraction of this estimate has been reported in existing crystal databases. Therefore, the study takes advantage of the reliable, inexpensive and rapid opportunity offered by deep generative modeling for accelerating the search for unknown perovskites. In the process of making such findings, an inverse design-modeling scheme is resolved, which aims at assimilating deterministic target properties with their corresponding perovskite structure. The inverse design pipeline is architectured by combining a generative Variational AutoEncoder (VAE) model with Target-Learning (TL) feed-forward neural networks to form the TL-VAE perovskite generator, thereby making the complete modeling process semi-supervisory. The TL feed-forward neural network model serves the purpose of organizing the non-linear latent space of the VAE model and further assists in isolating deterministic target properties that are of interest to the core objective of the study. The property to be target-learned in the latent space is the formation energy, which is a crucial indicator for calibrating perovskite stability. The results report the discovery of promising new perovskite candidates, which are unique and polymorphic material variants. Upon conducting Density Functional Theory (DFT) validation on the newly identified perovskites, candidates that undergo full geometrical relaxation are recommended for further investigation and/or synthesization. In conclusion, the study demonstrates the efficacy of the inverse design TL-VAE model for the generation of stable ternary perovskites.

Artificial neural network modelling for predicting classical air pressure profile curves in building drainage systems

Water trap seal loss can lead to the ingress of foul odours from the sewer network carrying harmful pathogens which can affect the health of building occupants. This loss is due to excessive negative or positive pressures within the system as a consequence of rapidly changing flow due to the unsteady nature of air and water flow discharges from appliances. This pressure regime within the system is often represented by a pressure profile along the height of the main vertical stack in the Building Drainage System (BDS). Experimental data from peer-reviewed literature and data from a unique 34-storey drainage test rig have been used as pressure profile data (Test data) for an Artificial Neural Network (ANN) model. Discharge input height (storey number) and discharge flow rate are considered to be the two independent input parameters and the pressure along the vertical stack is considered to be the output. In this work, both a Feed Forward and Back Propagation (FFBP) ANN model and a Radial Basis Function (RBF) ANN model have been used to train the algorithm. The work has confirmed the applicability of the FFBP-ANN model for steady two-phase fluid flow phenomena in BDS and allows for the prediction of pressures in a system for which no pressure data exists, by the prediction of modelled weights, based only on its physical and flow characteristics. Practical Application: Of great concern to designers of building drainage systems (BDS) is the control of pressure fluctuations within the system to prevent water trap seal loss. Prediction of a pressure profile based on the characteristics of the building – e.g. height, location of water discharge, discharge flow rate and ventilation type, would therefore increase confidence in designs, particularly in tall buildings. The work presented here addresses, for the first time, the applications of two ANN models for predicting the pressure profile in the BDS vertical stack of multi-storey buildings.

Reliable composite fault diagnosis of hydraulic systems based on linear discriminant analysis and multi-output hybrid kernel extreme learning machine

With increasingly stringent in requirements on the reliability and safety of hydraulic systems, data-driven fault diagnosis has emerged as a popular area of research. Hydraulic systems may have multiple failure modes, and accurately diagnosing compound failures in multi-component systems is a daunting task. In this paper, a method of multi-output classification by combining linear discriminant analysis (LDA) with the hybrid kernel extreme learning machine (HKELM) is proposed to diagnose compound faults in hydraulic systems. Data selection based on LDA is used in place of expert knowledge to screen out sensitive channels of each component from multi-channel signals. The multi-output strategy is embedded into the HKELM, which can simultaneously output the fault status of multiple components to diagnose the health of the system. An improved Hamming loss is also proposed to evaluate the total error in the multi-output classification because it has greater applicative relevance than classification accuracy. The results of experiments show that the proposed method can diagnose composite faults in multi-component systems with an accuracy higher than 99.5% and an error of only 0.20% on a dataset of hydraulic systems. As a shallow feed-forward network model, it can be used for real-time fault diagnosis due to its efficiency.

Stochastic capacity analysis for a distributed connected automated vehicle virtual car-following control strategy

Capacity analysis of the pure connected automated vehicle (CAV) traffic remains a challenging problem due to the high-dimensional factors involved in the control design. Especially, the communication loss and communication topology greatly impact the headway variation of CAVs and hence capacity with stochastic properties. This study provides a stochastic framework to mathematically derive multiple factors’ impact including free-flow speed, control gains, communication loss, and traffic arrival pattern on the pure CAV traffic capacity based on a virtual car-following control strategy targeting a single-lane highway and merging section. To begin with, we first mathematically derive the stochastic capacity for a single-lane highway based on a multi-predecessor-based linear feedback and feedforward car-following model for generic stochastic communication loss models. For a further illustration, a detailed analysis is conducted based on a well-known Signal-to-Interference-plus-Noise Ratio (SINR) communication loss model. We then extend the derivation to a merging section by a virtual car-following concept considering traffic arrival pattern’s stochasticity. Numerical sensitivity analyses have been conducted to systematically evaluate the impact of multiple factors mentioned. As the result indicated, the stochastic communication loss and traffic arrival pattern do have a significant impact on the pure CAV traffic capacity of the above scenarios.

Performance analysis of sentiment classification based on deep learning methods

When it comes to natural language processing, the textual differentiation task is one of the classical and important research problems. Recently, the deep learning model has increasingly become one of the main methods to solve text classification problems. Common deep learning text classification models are convolutional neural networks (CNN), recurrent neural networks (RNN), the BERT model. For comparing the manifestation of various deep learning models in textual differentiation tasks horizontally, the thesis tests the classification accuracy of different deep learning models under the same experimental configuration. The experimental results show that using pre-trained word vectors helps to improve the classification accuracy of deep learning models. In addition, the reasonable design of more complex and larger deep learning models is helpful in enhancing the study capability of the specific model on text data. The experimental results indicate that the text classification model using pre-trained word vectors could gain higher accuracy than the model without pre-trained word vectors. In addition, in the comparison experiment of feedforward neural network (FNN), CNN, RNN and BERT model, BERT model performs best, and the text classification accuracy reaches 0.9232. Compared with a 1-layer FNN, the accuracy rate is increased by about 16%.

Data-driven hedging of stock index options via deep learning

We develop deep learning models to learn the hedge ratio for S&P500 index options from options data. We compare different combinations of features and show that with sufficient training data, a feedforward neural network model with time to maturity, the Black-Scholes delta and market sentiment as inputs performs the best in the out-of-sample test under daily hedging. This model significantly outperforms delta hedging and a data-driven hedging model. Our results also demonstrate the importance of market sentiment for hedging.

Stable invariant models via Koopman spectra

Weight-tied models have attracted attention in the modern development of neural networks. The deep equilibrium model (DEQ) represents infinitely deep neural networks with weight-tying, and recent studies have shown the potential of this type of approach. DEQs are needed to iteratively solve root-finding problems in training and are built on the assumption that the underlying dynamics determined by the models converge to a fixed point. In this paper, we present the stable invariant model (SIM), a new class of deep models that in principle approximates DEQs under stability and extends the dynamics to more general ones converging to an invariant set (not restricted in a fixed point). The key ingredient in deriving SIMs is a representation of the dynamics with the spectra of the Koopman and Perron–Frobenius operators. This perspective approximately reveals stable dynamics with DEQs and then derives two variants of SIMs. We also propose an implementation of SIMs that can be learned in the same way as feedforward models. We illustrate the empirical performance of SIMs with experiments and demonstrate that SIMs achieve comparative or superior performance against DEQs in several learning tasks.

Adaptive output feedback control of uncertain nonlinear systems with input saturation

AbstractThis article concerns with the tracking control problem for a class of control‐affine nonlinear systems subject to input saturation, parametric uncertainties and unmodeled uncertainties. The proposed scheme consists of a nested‐saturation‐function‐based controller integrated with feedforward model compensation. A saturated linear extended state observer (SLESO) and parameter adaptation law are constructed to compensate for the unmodeled uncertainties and parametric uncertainties, respectively. By using these mechanisms, the steady‐state tracking performance can be guaranteed through uncertainty compensation. Additionally, the issue of conservativeness in the input saturation problem is effectively addressed through online‐updating the available unsaturated region, which improves transient performance. The proposed scheme results in asymptotic stability when unmodeled uncertainties are constants and ultimate boundedness when unmodeled uncertainties are time‐varying. Finally, simulation studies are presented to demonstrate the effectiveness of the proposed approach.

Modelling the Employment Rate in Türkiye Using Machine Learning Methods

Employment rate is a crucial economic indicator that measures the percentage of the working-age population that is employed. Considering this importance, the employment rate of Türkiye is modelled in this work. The employment rate, gross domestic product, gross fixed capital formation, government expenditures, export and import data of the 1991-2021 period are taken from the World bank database and then the seasonal-trend decomposition of these data are performed in EViews software. After inspecting the nonlinearity of the data, a nonlinear machine learning model namely a feedforward artificial neural network model is developed in Python programming language for the modelling of the employment rate dependent on the gross domestic product, gross fixed capital formation, government expenditures, export and import data. The 70% of the available data is used as the training data and the developed feedforward artificial neural network is trained with a successful convergence despite the low number of samples thanks to its optimal structure. Then the remaining 30% of the available data are utilized as the test data. The actual employment rate data and the result of the developed feedforward neural network model are plotted on the same axis pair which show continuous overlap in a wide range. Apart from the visual inspection of the results of the developed model, the performance metrics are also calculated in Python programming language such as the mean absolute percentage error, coefficient of determination, root mean square error and the mean absolute error. The performance metrics of the results of the developed model also verify the high accuracy of the model. It is argued that the developed feedforward neural network model can also be used for the situations where low number of samples need to be modelled thanks to the optimal structure. Keywords: Employment rate, modelling, machine learning, artificial neural networks, estimation.

Modeling hysteresis in expanded polystyrene foams under compressive loads using feed-forward neural networks

Expanded polystyrene foams are widely used materials for various applications in engineering, including their use for protective designs. For this type of application, in engineering analysis and design, it is required to know the mechanical response to compression of this type of material, since energy parameters that support the analysis of the effectiveness of a design are derived from it. One of these parameters is strain hysteresis, through which it is possible to know how capable a material is of absorbing energy. The modeling and prediction of this parameter is a challenge from the analysis point of view. This contribution presents a method based on feed-forward artificial neural network models that address a modeling approach to derive this parameter from the mechanical response of expanded polystyrene foam. From this, models are constructed that can predict the response of such material to various density and loading rate conditions. The best of a total of 30 neural network models, which are capable of deriving energy parameters such as hysteresis, is chosen. The results show that this approach is valid for the deformation energy analysis of expanded polystyrene foams since results consistent with the material phenomenology and errors of less than 3% with respect to experimental data are obtained.