Search results for: neural network models

Authors: B. R. Campomanes-Alvarez, P. Quiros, B. Fernandez

Abstract:

Automatic Speech Recognition (ASR) is a machine-based process of decoding and transcribing oral speech. A typical ASR system receives acoustic input from a speaker or an audio file, analyzes it using algorithms, and produces an output in the form of a text. Some speech recognition systems use Hidden Markov Models (HMMs) to deal with the temporal variability of speech and Gaussian Mixture Models (GMMs) to determine how well each state of each HMM fits a short window of frames of coefficients that represents the acoustic input. Another way to evaluate the fit is to use a feed-forward neural network that takes several frames of coefficients as input and produces posterior probabilities over HMM states as output. Deep neural networks (DNNs) that have many hidden layers and are trained using new methods have been shown to outperform GMMs on a variety of speech recognition systems. Acoustic models for state-of-the-art ASR systems are usually training on massive amounts of data. However, audio files with their corresponding transcriptions can be difficult to obtain, especially in the Spanish language. Hence, in the case of these low-resource scenarios, building an ASR model is considered as a complex task due to the lack of labeled data, resulting in an under-trained system. Semi-supervised learning approaches arise as necessary tasks given the high cost of transcribing audio data. The main goal of this proposal is to develop a procedure based on acoustic semi-supervised learning for Spanish ASR systems by using DNNs. This semi-supervised learning approach consists of: (a) Training a seed ASR model with a DNN using a set of audios and their respective transcriptions. A DNN with a one-hidden-layer network was initialized; increasing the number of hidden layers in training, to a five. A refinement, which consisted of the weight matrix plus bias term and a Stochastic Gradient Descent (SGD) training were also performed. The objective function was the cross-entropy criterion. (b) Decoding/testing a set of unlabeled data with the obtained seed model. (c) Selecting a suitable subset of the validated data to retrain the seed model, thereby improving its performance on the target test set. To choose the most precise transcriptions, three confidence scores or metrics, regarding the lattice concept (based on the graph cost, the acoustic cost and a combination of both), was performed as selection technique. The performance of the ASR system will be calculated by means of the Word Error Rate (WER). The test dataset was renewed in order to extract the new transcriptions added to the training dataset. Some experiments were carried out in order to select the best ASR results. A comparison between a GMM-based model without retraining and the DNN proposed system was also made under the same conditions. Results showed that the semi-supervised ASR-model based on DNNs outperformed the GMM-model, in terms of WER, in all tested cases. The best result obtained an improvement of 6% relative WER. Hence, these promising results suggest that the proposed technique could be suitable for building ASR models in low-resource environments.

Keywords: automatic speech recognition, deep neural networks, machine learning, semi-supervised learning

Procedia PDF Downloads 341

Authors: Khalid Salhi, El Miloud Jaara, Mohammed Talibi Alaoui

Abstract:

In this paper, we present a neural approach for unsupervised natural color-texture image segmentation, which is based on both Kohonen maps and mathematical morphology, using a combination of the texture and the image color information of the image, namely, the fractal features based on fractal dimension are selected to present the information texture, and the color features presented in RGB color space. These features are then used to train the network Kohonen, which will be represented by the underlying probability density function, the segmentation of this map is made by morphological watershed transformation. The performance of our color-texture segmentation approach is compared first, to color-based methods or texture-based methods only, and then to k-means method.

Keywords: segmentation, color-texture, neural networks, fractal, watershed

Procedia PDF Downloads 349

Authors: Apoorva Vinod, Archana Mathur, Snehanshu Saha

Abstract:

Though there exists a plethora of Activation Functions (AFs) used in single and multiple hidden layer Neural Networks (NN), their behavior always raised curiosity, whether used in combination or singly. The popular AFs –Sigmoid, ReLU, and Tanh–have performed prominently well for shallow and deep architectures. Most of the time, AFs are used singly in multi-layered NN, and, to the best of our knowledge, their performance is never studied and analyzed deeply when used in combination. In this manuscript, we experiment with multi-layered NN architecture (both on shallow and deep architectures; Convolutional NN and VGG16) and investigate how well the network responds to using two different AFs (Sigmoid-Tanh, Tanh-ReLU, ReLU-Sigmoid) used alternately against a traditional, single (Sigmoid-Sigmoid, Tanh-Tanh, ReLUReLU) combination. Our results show that using two different AFs, the network achieves better accuracy, substantially lower loss, and faster convergence on 4 computer vision (CV) and 15 Non-CV (NCV) datasets. When using different AFs, not only was the accuracy greater by 6-7%, but we also accomplished convergence twice as fast. We present a case study to investigate the probability of networks suffering vanishing and exploding gradients when using two different AFs. Additionally, we theoretically showed that a composition of two or more AFs satisfies Universal Approximation Theorem (UAT).

Keywords: activation function, universal approximation function, neural networks, convergence

Procedia PDF Downloads 160

Authors: Semachew M. Kassa, Africa M Geremew, Tezera F. Azmatch, Nandyala Darga Kumar

Abstract:

Frequency ratio (FR) and analytical hierarchy process (AHP) methods are developed based on past landslide failure points to identify the landslide susceptibility mapping because landslides can seriously harm both the environment and society. However, it is still difficult to select the most efficient method and correctly identify the main driving factors for particular regions. In this study, we used fourteen landslide conditioning factors (LCFs) and five soft computing algorithms, including Random Forest (RF), Support Vector Machine (SVM), Logistic Regression (LR), Artificial Neural Network (ANN), and Naïve Bayes (NB), to predict the landslide susceptibility at 12.5 m spatial scale. The performance of the RF (F1-score: 0.88, AUC: 0.94), ANN (F1-score: 0.85, AUC: 0.92), and SVM (F1-score: 0.82, AUC: 0.86) methods was significantly better than the LR (F1-score: 0.75, AUC: 0.76) and NB (F1-score: 0.73, AUC: 0.75) method, according to the classification results based on inventory landslide points. The findings also showed that around 35% of the study region was made up of places with high and very high landslide risk (susceptibility greater than 0.5). The very high-risk locations were primarily found in the western and southeastern regions, and all five models showed good agreement and similar geographic distribution patterns in landslide susceptibility. The towns with the highest landslide risk include Amhara Saint Town's western part, the Northern part, and St. Gebreal Church villages, with mean susceptibility values greater than 0.5. However, rainfall, distance to road, and slope were typically among the top leading factors for most villages. The primary contributing factors to landslide vulnerability were slightly varied for the five models. Decision-makers and policy planners can use the information from our study to make informed decisions and establish policies. It also suggests that various places should take different safeguards to reduce or prevent serious damage from landslide events.

Keywords: artificial neural network, logistic regression, landslide susceptibility, naïve Bayes, random forest, support vector machine

Procedia PDF Downloads 84

Authors: Masood Varshosaz, Kamyar Hasanpour

Abstract:

In recent years, we have seen growing interest in recognizing humans in drone images for post-disaster search and rescue operations. Deep learning algorithms have shown great promise in this area, but they often require large amounts of labeled data to train the models. To keep the data acquisition cost low, augmentation techniques can be used to create additional data from existing images. There are many techniques of such that can help generate variations of an original image to improve the performance of deep learning algorithms. While data augmentation is potentially assumed to improve the accuracy and robustness of the models, it is important to ensure that the performance gains are not outweighed by the additional computational cost or complexity of implementing the techniques. To this end, it is important to evaluate the impact of data augmentation on the performance of the deep learning models. In this paper, we evaluated the most currently available 2D data augmentation techniques on a standard convolutional network which was trained for recognizing humans in drone images. The techniques include rotation, scaling, random cropping, flipping, shifting, and their combination. The results showed that the augmented models perform 1-3% better compared to a base network. However, as the augmented images only contain the human parts already visible in the original images, a new data augmentation approach is needed to include the invisible parts of the human body. Thus, we suggest a new method that employs simulated 3D human models to generate new data for training the network.

Keywords: human recognition, deep learning, drones, disaster mitigation

Procedia PDF Downloads 96

Authors: Jameela Ali, Hamid A. Jalab, Loay E. George, Abdul Rahim Ahmad, Azizah Suliman, Karim Al-Jashamy

Abstract:

Red blood cells (RBC) are the most common types of blood cells and are the most intensively studied in cell biology. The lack of RBCs is a condition in which the amount of hemoglobin level is lower than normal and is referred to as “anemia”. Abnormalities in RBCs will affect the exchange of oxygen. This paper presents a comparative study for various techniques for classifying the RBCs as normal, or abnormal (anemic) using WEKA. WEKA is an open source consists of different machine learning algorithms for data mining applications. The algorithm tested are Radial Basis Function neural network, Support vector machine, and K-Nearest Neighbors algorithm. Two sets of combined features were utilized for classification of blood cells images. The first set, exclusively consist of geometrical features, was used to identify whether the tested blood cell has a spherical shape or non-spherical cells. While the second set, consist mainly of textural features was used to recognize the types of the spherical cells. We have provided an evaluation based on applying these classification methods to our RBCs image dataset which were obtained from Serdang Hospital-alaysia, and measuring the accuracy of test results. The best achieved classification rates are 97%, 98%, and 79% for Support vector machines, Radial Basis Function neural network, and K-Nearest Neighbors algorithm respectively.

Keywords: K-nearest neighbors algorithm, radial basis function neural network, red blood cells, support vector machine

Procedia PDF Downloads 411

Authors: Despoina Mouratidis, Katia Lida Kermanidis

Abstract:

Machine translation (MT) is a very important task in Natural Language Processing (NLP). MT evaluation is crucial in MT development, as it constitutes the means to assess the success of an MT system, and also helps improve its performance. Several methods have been proposed for the evaluation of (MT) systems. Some of the most popular ones in automatic MT evaluation are score-based, such as the BLEU score, and others are based on lexical similarity or syntactic similarity between the MT outputs and the reference involving higher-level information like part of speech tagging (POS). This paper presents a language-independent machine learning framework for classifying pairwise translations. This framework uses vector representations of two machine-produced translations, one from a statistical machine translation model (SMT) and one from a neural machine translation model (NMT). The vector representations consist of automatically extracted word embeddings and string-like language-independent features. These vector representations used as an input to a multi-layer neural network (NN) that models the similarity between each MT output and the reference, as well as between the two MT outputs. To evaluate the proposed approach, a professional translation and a "ground-truth" annotation are used. The parallel corpora used are English-Greek (EN-GR) and English-Italian (EN-IT), in the educational domain and of informal genres (video lecture subtitles, course forum text, etc.) that are difficult to be reliably translated. They have tested three basic deep learning (DL) architectures to this schema: (i) fully-connected dense, (ii) Convolutional Neural Network (CNN), and (iii) Long Short-Term Memory (LSTM). Experiments show that all tested architectures achieved better results when compared against those of some of the well-known basic approaches, such as Random Forest (RF) and Support Vector Machine (SVM). Better accuracy results are obtained when LSTM layers are used in our schema. In terms of a balance between the results, better accuracy results are obtained when dense layers are used. The reason for this is that the model correctly classifies more sentences of the minority class (SMT). For a more integrated analysis of the accuracy results, a qualitative linguistic analysis is carried out. In this context, problems have been identified about some figures of speech, as the metaphors, or about certain linguistic phenomena, such as per etymology: paronyms. It is quite interesting to find out why all the classifiers led to worse accuracy results in Italian as compared to Greek, taking into account that the linguistic features employed are language independent.

Keywords: machine learning, machine translation evaluation, neural network architecture, pairwise classification

Procedia PDF Downloads 133

Authors: Hamed Ouattara, Pierre Duthon, Frédéric Bernardin, Omar Ait Aider, Pascal Salmane

Abstract:

In this article, we present three neural network architectures for real-time classification of weather conditions (sunny, rainy, snowy, foggy) from images. Inspired by recent advances in style transfer, two of these architectures -Truncated ResNet50 and Truncated ResNet50 with Gram Matrix and Attention- surpass the state of the art and demonstrate re-markable generalization capability on several public databases, including Kaggle (2000 images), Kaggle 850 images, MWI (1996 images) [1], and Image2Weather [2]. Although developed for weather detection, these architectures are also suitable for other appearance-based classification tasks, such as animal species recognition, texture classification, disease detection in medical images, and industrial defect identification. We illustrate these applications in the section “Applications of Our Models to Other Tasks” with the “SIIM-ISIC Melanoma Classification Challenge 2020” [3].

Keywords: weather simulation, weather measurement, weather classification, weather detection, style transfer, Pix2Pix, CycleGAN, CUT, neural style transfer

Procedia PDF Downloads 12

Authors: Ja-Keoung Koo, Kensuke Nakamura, Hyohun Kim, Dongwha Shin, Yeonseok Kim, Ji-Su Ahn, Byung-Woo Hong

Abstract:

The machine learning techniques based on a convolutional neural network (CNN) have been actively developed and successfully applied to a variety of image analysis tasks including reconstruction, noise reduction, resolution enhancement, segmentation, motion estimation, object recognition. The classical visual information processing that ranges from low level tasks to high level ones has been widely developed in the deep learning framework. It is generally considered as a challenging problem to derive visual interpretation from high dimensional imagery data. A CNN is a class of feed-forward artificial neural network that usually consists of deep layers the connections of which are established by a series of non-linear operations. The CNN architecture is known to be shift invariant due to its shared weights and translation invariance characteristics. However, it is often computationally intractable to optimize the network in particular with a large number of convolution layers due to a large number of unknowns to be optimized with respect to the training set that is generally required to be large enough to effectively generalize the model under consideration. It is also necessary to limit the size of convolution kernels due to the computational expense despite of the recent development of effective parallel processing machinery, which leads to the use of the constantly small size of the convolution kernels throughout the deep CNN architecture. However, it is often desired to consider different scales in the analysis of visual features at different layers in the network. Thus, we propose a CNN model where different sizes of the convolution kernels are applied at each layer based on the random projection. We apply random filters with varying sizes and associate the filter responses with scalar weights that correspond to the standard deviation of the random filters. We are allowed to use large number of random filters with the cost of one scalar unknown for each filter. The computational cost in the back-propagation procedure does not increase with the larger size of the filters even though the additional computational cost is required in the computation of convolution in the feed-forward procedure. The use of random kernels with varying sizes allows to effectively analyze image features at multiple scales leading to a better generalization. The robustness and effectiveness of the proposed CNN based on random kernels are demonstrated by numerical experiments where the quantitative comparison of the well-known CNN architectures and our models that simply replace the convolution kernels with the random filters is performed. The experimental results indicate that our model achieves better performance with less number of unknown weights. The proposed algorithm has a high potential in the application of a variety of visual tasks based on the CNN framework. Acknowledgement—This work was supported by the MISP (Ministry of Science and ICT), Korea, under the National Program for Excellence in SW (20170001000011001) supervised by IITP, and NRF-2014R1A2A1A11051941, NRF2017R1A2B4006023.

Keywords: deep learning, convolutional neural network, random kernel, random projection, dimensionality reduction, object recognition

Procedia PDF Downloads 291

Authors: Hadi Rouhi Belvirdi, Davoud Beheshtizadeh

Abstract:

The increasing demand for sustainable construction materials has led to a growing interest in high-strength recycled concrete (HSRC). Utilizing recycled materials not only reduces waste but also minimizes the depletion of natural resources. This study explores the application of artificial intelligence (AI) techniques to model and predict the properties of HSRC. In the past two decades, the production levels in various industries and, consequently, the amount of waste have increased significantly. Continuing this trend will undoubtedly cause irreparable damage to the environment. For this reason, engineers have been constantly seeking practical solutions for recycling industrial waste in recent years. This research utilized the results of the compressive strength of 90-day high-strength recycled concrete. The method for creating recycled concrete involved replacing sand with crushed glass and using glass powder instead of cement. Subsequently, a feedforward artificial neural network was employed to model the compressive strength results for 90 days. The regression and error values obtained indicate that this network is suitable for modeling the compressive strength data.

Keywords: high-strength recycled concrete, feedforward artificial neural network, regression, construction materials

Procedia PDF Downloads 17

Authors: Martin K. Steiger, Lukas Heisler, Hans-Georg Brachtendorf

Abstract:

Deep neural networks and their variants form the backbone of many AI applications. Based on the so-called residual networks, a continuous formulation of such models as ordinary differential equations (ODEs) has proven advantageous since different techniques may be applied that significantly increase the learning speed and enable controlled trade-offs with the resulting error at the same time. For the evaluation of such models, high-performance numerical differential equation solvers are used, which also provide the gradients required for training. However, whether classical gradient-based methods are even applicable or which one yields the best results has not been discussed yet. This paper aims to redeem this situation by providing empirical results for different applications.

Keywords: deep neural networks, gradient-based learning, image processing, ordinary differential equation networks

Procedia PDF Downloads 170

Authors: Christian H. Sanabria-Montaña, Rodrigo Huerta-Quintanilla

Abstract:

A lattice network is a special type of network in which all nodes have the same number of links, and its boundary conditions are periodic. The most basic lattice network is the ring, a one-dimensional network with periodic border conditions. In contrast, the Cartesian product of d rings forms a d-dimensional lattice network. An analytical expression currently exists for the clustering coefficient in this type of network, but the theoretical value is valid only up to certain connectivity value; in other words, the analytical expression is incomplete. Here we obtain analytically the clustering coefficient expression in d-dimensional lattice networks for any link density. Our analytical results show that the clustering coefficient for a lattice network with density of links that tend to 1, leads to the value of the clustering coefficient of a fully connected network. We developed a model on criminology in which the generalized clustering coefficient expression is applied. The model states that delinquents learn the know-how of crime business by sharing knowledge, directly or indirectly, with their friends of the gang. This generalization shed light on the network properties, which is important to develop new models in different fields where network structure plays an important role in the system dynamic, such as criminology, evolutionary game theory, econophysics, among others.

Keywords: clustering coefficient, criminology, generalized, regular network d-dimensional

Procedia PDF Downloads 412

Authors: Muhammad Atif, Cang Yan

Abstract:

The enhancement of low-light images is a significant area of study aimed at enhancing the quality of captured images in challenging lighting environments. Recently, methods based on convolutional neural networks (CNN) have gained prominence as they offer state-of-the-art performance. However, many approaches based on CNN rely on increasing the size and complexity of the neural network. In this study, we propose an alternative method for improving low-light images using an autoencoder-based multiscale knowledge transfer model. Our method leverages the power of three autoencoders, where the encoders of the first two autoencoders are directly connected to the decoder of the third autoencoder. Additionally, the decoder of the first two autoencoders is connected to the encoder of the third autoencoder. This architecture enables effective knowledge transfer, allowing the third autoencoder to learn and benefit from the enhanced knowledge extracted by the first two autoencoders. We further integrate the proposed model into the PIX to PIX GAN framework. By integrating our proposed model as the generator in the GAN framework, we aim to produce enhanced images that not only exhibit improved visual quality but also possess a more authentic and realistic appearance. These experimental results, both qualitative and quantitative, show that our method is better than the state-of-the-art methodologies.

Keywords: low light image enhancement, deep learning, convolutional neural network, image processing

Procedia PDF Downloads 85

Authors: Lizhi Ma, Dan Liu

Abstract:

Pipeline parallelism has been widely used to accelerate distributed deep learning to alleviate GPU memory bottlenecks and to ensure that models can be trained and deployed smoothly under limited graphics memory conditions. However, in highly heterogeneous distributed clusters, traditional model partitioning methods are not able to achieve load balancing. The overlap of communication and computation is also a big challenge. In this paper, HePipe is proposed, an efficient pipeline parallel training method for highly heterogeneous clusters. According to the characteristics of the neural network model pipeline training task, oriented to the 2-level heterogeneous cluster computing topology, a training method based on the 2-level stage division of neural network modeling and partitioning is designed to improve the parallelism. Additionally, a multi-forward 1F1B scheduling strategy is designed to accelerate the training time of each stage by executing the computation units in advance to maximize the overlap between the forward propagation communication and backward propagation computation. Finally, a dynamic recomputation strategy based on task memory requirement prediction is proposed to improve the fitness ratio of task and memory, which improves the throughput of the cluster and solves the memory shortfall problem caused by memory differences in heterogeneous clusters. The empirical results show that HePipe improves the training speed by 1.6×−2.2× over the existing asynchronous pipeline baselines.

Keywords: pipeline parallelism, heterogeneous cluster, model training, 2-level stage partitioning

Procedia PDF Downloads 19

Authors: Elia Favarelli, Enrico Testi, Andrea Giorgetti

Abstract:

Autonomous structural health monitoring (SHM) of many structures and bridges became a topic of paramount importance for maintenance purposes and safety reasons. This paper proposes a set of machine learning (ML) tools to perform automatic feature selection and detection of anomalies in a bridge from vibrational data and compare different feature extraction schemes to increase the accuracy and reduce the amount of data collected. As a case study, the Z-24 bridge is considered because of the extensive database of accelerometric data in both standard and damaged conditions. The proposed framework starts from the first four fundamental frequencies extracted through operational modal analysis (OMA) and clustering, followed by density-based time-domain filtering (tracking). The fundamental frequencies extracted are then fed to a dimensionality reduction block implemented through two different approaches: feature selection (intelligent multiplexer) that tries to estimate the most reliable frequencies based on the evaluation of some statistical features (i.e., mean value, variance, kurtosis), and feature extraction (auto-associative neural network (ANN)) that combine the fundamental frequencies to extract new damage sensitive features in a low dimensional feature space. Finally, one class classifier (OCC) algorithms perform anomaly detection, trained with standard condition points, and tested with normal and anomaly ones. In particular, a new anomaly detector strategy is proposed, namely one class classifier neural network two (OCCNN2), which exploit the classification capability of standard classifiers in an anomaly detection problem, finding the standard class (the boundary of the features space in normal operating conditions) through a two-step approach: coarse and fine boundary estimation. The coarse estimation uses classics OCC techniques, while the fine estimation is performed through a feedforward neural network (NN) trained that exploits the boundaries estimated in the coarse step. The detection algorithms vare then compared with known methods based on principal component analysis (PCA), kernel principal component analysis (KPCA), and auto-associative neural network (ANN). In many cases, the proposed solution increases the performance with respect to the standard OCC algorithms in terms of F1 score and accuracy. In particular, by evaluating the correct features, the anomaly can be detected with accuracy and an F1 score greater than 96% with the proposed method.

Keywords: anomaly detection, frequencies selection, modal analysis, neural network, sensor network, structural health monitoring, vibration measurement

Procedia PDF Downloads 124

Authors: Obe Olumide Olayinka, Victor Balanica, Eugen Neagoe

Abstract:

The effects of hypertension are often lethal thus its early detection and prevention is very important for everybody. In this paper, a neural network (NN) model was developed and trained based on a dataset of hypertension causative parameters in order to forecast the likelihood of occurrence of hypertension in patients. Our research goal was to analyze the potential of the presented NN to predict, for a period of time, the risk of hypertension or the risk of developing this disease for patients that are or not currently hypertensive. The results of the analysis for a given patient can support doctors in taking pro-active measures for averting the occurrence of hypertension such as recommendations regarding the patient behavior in order to lower his hypertension risk. Moreover, the paper envisages a set of three example scenarios in order to determine the age when the patient becomes hypertensive, i.e. determine the threshold for hypertensive age, to analyze what happens if the threshold hypertensive age is set to a certain age and the weight of the patient if being varied, and, to set the ideal weight for the patient and analyze what happens with the threshold of hypertensive age.

Keywords: neural network, hypertension, data set, training set, supervised learning

Procedia PDF Downloads 394

Authors: Zheng Zhihao

Abstract:

Recently, Knowledge Graph Framework Network (KGCN) has developed powerful capabilities in knowledge representation and reasoning tasks. However, traditional KGCN often uses a fixed weight mechanism when aggregating information, failing to make full use of rich structural information, resulting in a certain expression ability of node representation, and easily causing over-smoothing problems. In order to solve these challenges, the paper proposes an new graph neural network model called KGCN-STAR (Knowledge Graph Convolutional Network with Structural Adaptive Receptive Fields). This model dynamically adjusts the perception of each node by introducing a structural adaptive receptive field. wild range, and a subgraph aggregator is designed to capture local structural information more effectively. Experimental results show that KGCN-STAR shows significant performance improvement on multiple knowledge graph data sets, especially showing considerable capabilities in the task of representation learning of complex structures.

Keywords: knowledge graph, graph neural networks, structural adaptive receptive fields, information aggregation

Procedia PDF Downloads 35

Authors: S. Dhib, S. A. Addouche, T. Loukil, A. Elmhamedi

Abstract:

In this paper, a reverse supply chain network is investigated for a decision making. This decision is surrounded by complex flows of returned products, due to the increasing quantity, the type of returned products and the variety of recovery option products (reuse, recycling, and refurbishment). The most important problem in the reverse logistic network (RLN) is to orient returned products to the suitable type of recovery option. However, returned products orientations from collect sources to the recovery disposition have not well considered in performance model. In this study, we propose a performance model for designing a network configuration on reverse logistics. Conceptual and analytical models are developed with taking into account operational, economic and environmental factors on designing network.

Keywords: reverse logistics, network design, performance model, open loop configuration

Procedia PDF Downloads 435

Authors: Tian Lan, Yixiang Wang, Wenxin Tai, Yilan Lyu, Zufeng Wu

Abstract:

The monaural speech enhancement task in the time-frequencydomain has a myriad of approaches, with the stacked con-volutional neural network (CNN) demonstrating superiorability in feature extraction and selection. However, usingstacked single convolutions method limits feature represen-tation capability and generalization ability. In order to solvethe aforementioned problem, we propose an attention-basedadaptive convolutional network that integrates the multi-scale convolutional operations into a operation-specific blockvia input dependent attention to adapt to complex auditoryscenes. In addition, we introduce a two-stage progressivelearning method to enlarge the receptive field without a dra-matic increase in computation burden. We conduct a series ofexperiments based on the TIMIT corpus, and the experimen-tal results prove that our proposed model is better than thestate-of-art models on all metrics.

Keywords: speech enhancement, adaptive convolu-tion, progressive learning, time-frequency domain

Procedia PDF Downloads 124

Authors: Vuk M. Popovic, Dunja D. Popovic

Abstract:

Software maintenance phase is started once a software project has been developed and delivered. After that, any modification to it corresponds to maintenance. Software maintenance involves modifications to keep a software project usable in a changed or a changing environment, to correct discovered faults, and modifications, and to improve performance or maintainability. Software maintenance and management of software maintenance are recognized as two most important and most expensive processes in a life of a software product. This research is basing the prediction of maintenance, on risks and time evaluation, and using them as data sets for working with neural networks. The aim of this paper is to provide support to project maintenance managers. They will be able to pass the issues planned for the next software-service-patch to the experts, for risk and working time evaluation, and afterward to put all data to neural networks in order to get software maintenance prediction. This process will lead to the more accurate prediction of the working hours needed for the software-service-patch, which will eventually lead to better planning of budget for the software maintenance projects.

Keywords: laboratory information system, maintenance engineering, neural networks, software maintenance, software maintenance costs

Procedia PDF Downloads 360

Authors: Seunghee Park, Junkyeong Kim, Eun-Seok Shin, Sang-Hun Han

Abstract:

In this study, two kinds of nondestructive evaluation (NDE) techniques (rebound hardness and ultrasonic pulse velocity methods) are investigated for the effective maintenance of underwater concrete structures. A new methodology to estimate the underwater concrete strengths more effectively, named “artificial neural network (ANN) – based concrete strength estimation with the combination of rebound hardness and ultrasonic pulse velocity methods” is proposed and verified throughout a series of experimental works.

Keywords: underwater concrete, rebound hardness, Schmidt hammer, ultrasonic pulse velocity, ultrasonic sensor, artificial neural networks, ANN

Procedia PDF Downloads 533

Authors: P. Neeraj, Anurag Sharma, Harsukhpreet Singh

Abstract:

In current years, Brain-Computer Interface (BCI) scheme based on steady-state Visual Evoked Potential (SSVEP) have earned much consideration. This study tries to evolve an SSVEP based BCI scheme that can regulate any gadget mock-up in two unique positions ON and OFF. In this paper, two distinctive gleam frequencies in low-frequency part were utilized to evoke the SSVEPs and were shown on a Liquid Crystal Display (LCD) screen utilizing Lab View. Two stimuli shading, Yellow, and Blue were utilized to prepare the system in SSVEPs. The Electroencephalogram (EEG) signals recorded from the occipital part. Elements of the brain were separated by utilizing discrete wavelet Transform. A prominent system for multilayer system diverse Neural Network Algorithm (NNA), is utilized to characterize SSVEP signals. During training of the network with diverse calculation Regression plot results demonstrated that when Levenberg-Marquardt preparing calculation was utilized the exactness turns out to be 93.9%, which is superior to another training algorithm.

Keywords: brain computer interface, electroencephalography, steady-state visual evoked potential, wavelet transform, neural network

Procedia PDF Downloads 334

Authors: Supriya H. S., Chandrakala B. M.

Abstract:

Deep learning based models have been recently successful adoption for network traffic prediction. However, training a deep learning model for various prediction tasks is considered one of the critical tasks due to various reasons. This research work develops Multi-Layer Hybrid Network (MLHN) for network traffic prediction and analysis; MLHN comprises the three distinctive networks for handling the different inputs for custom feature extraction. Furthermore, an optimized and efficient parameter-tuning algorithm is introduced to enhance parameter learning. MLHN is evaluated considering the “Big Data Challenge” dataset considering the Mean Absolute Error, Root Mean Square Error and R^2as metrics; furthermore, MLHN efficiency is proved through comparison with a state-of-art approach.

Keywords: MLHN, network traffic prediction

Procedia PDF Downloads 90

Authors: Joselito Medina-Marin, Maria G. Serna-Diaz, Juan C. Seck-Tuoh-Mora, Norberto Hernandez-Romero, Irving Barragán-Vite

Abstract:

Wind speed forecasting is an important issue for planning wind power generation facilities. The accuracy in the wind speed prediction allows a good performance of wind turbines for electricity generation. A model based on artificial neural networks is presented in this work. A dataset with atmospheric information about air temperature, atmospheric pressure, wind direction, and wind speed in Pachuca, Hidalgo, México, was used to train the artificial neural network. The data was downloaded from the web page of the National Meteorological Service of the Mexican government. The records were gathered for three months, with time intervals of ten minutes. This dataset was used to develop an iterative algorithm to create 1,110 ANNs, with different configurations, starting from one to three hidden layers and every hidden layer with a number of neurons from 1 to 10. Each ANN was trained with the Levenberg-Marquardt backpropagation algorithm, which is used to learn the relationship between input and output values. The model with the best performance contains three hidden layers and 9, 6, and 5 neurons, respectively; and the coefficient of determination obtained was r²=0.9414, and the Root Mean Squared Error is 1.0559. In summary, the ANN approach is suitable to predict the wind speed in Pachuca City because the r² value denotes a good fitting of gathered records, and the obtained ANN model can be used in the planning of wind power generation grids.

Keywords: wind power generation, artificial neural networks, wind speed, coefficient of determination

Procedia PDF Downloads 125

Authors: Juan José Mesas, Luis Sainz

Abstract:

The motivation for the development of this paper lies in the need for energy networks to reduce losses, improve performance, optimize their operation and try to benefit from the interconnection capacity with other networks enabled for other energy carriers. These interconnections generate interdependencies between some energy networks and others, which requires suitable models and methods for their analysis. Traditionally, the modeling and study of energy networks have been carried out independently for each energy carrier. Thus, there are well-established models and methods for the steady-state analysis of electrical networks, gas networks, and thermal networks separately. What is intended is to extend and combine them adequately to be able to face in an integrated way the steady-state analysis of networks with multiple energy carriers. Firstly, the added value of multi-energy networks, their operation, and the basic principles that characterize them are explained. In addition, two current aspects of great relevance are exposed: the storage technologies and the coupling elements used to interconnect one energy network with another. Secondly, the characteristic equations of the different energy networks necessary to carry out the steady-state analysis are detailed. The electrical network, the natural gas network, and the thermal network of heat and cold are considered in this paper. After the presentation of the equations, a particular case of the steady-state analysis of a specific multi-energy network is studied. This network is represented graphically, the interconnections between the different energy carriers are described, their technical data are exposed and the equations that have previously been presented theoretically are formulated and developed. Finally, the two iterative numerical resolution methods considered in this paper are presented, as well as the resolution procedure and the results obtained. The pros and cons of the application of both methods are explained. It is verified that the results obtained for the electrical network (voltages in modulus and angle), the natural gas network (pressures), and the thermal network (mass flows and temperatures) are correct since they comply with the distribution, operation, consumption and technical characteristics of the multi-energy network under study.

Keywords: coupling elements, energy carriers, multi-energy networks, steady-state analysis

Procedia PDF Downloads 81

Authors: Shubhakar Gupta, Dhruv Prakash, Apoorv Mehta

Abstract:

In this project, we design a modular robot capable of using and switching between multiple methods of propulsion and classifying terrain, based on an Inertial Measurement Unit (IMU) input. We wanted to make a robot that is not only intelligent in its functioning but also versatile in its physical design. The advantage of a modular robot is that it can be designed to hold several movement-apparatuses, such as wheels, legs for a hexapod or a quadpod setup, propellers for underwater locomotion, and any other solution that may be needed. The robot takes roughness input from a gyroscope and an accelerometer in the IMU, and based on the terrain classification from an artificial neural network; it decides which method of propulsion would best optimize its movement. This provides the bot with adaptability over a set of terrains, which means it can optimize its locomotion on a terrain based on its roughness. A feature like this would be a great asset to have in autonomous exploration or research drones.

Keywords: modular robotics, terrain detection, terrain classification, neural network

Procedia PDF Downloads 147

Authors: M. Caramihai, Irina Severin

Abstract:

This paper deals with the design of an intelligent control structure for a bioprocess of Hansenula polymorpha yeast cultivation. The objective of the process control is to produce biomass in a desired physiological state. The work demonstrates that the designed Hybrid Control Techniques (HCT) are able to recognize specific evolution bioprocess trajectories using neural networks trained specifically for this purpose, in order to estimate the model parameters and to adjust the overall bioprocess evolution through an expert system and a fuzzy structure. The design of the control algorithm as well as its tuning through realistic simulations is presented. Taking into consideration the synergism of different paradigms like fuzzy logic, neural network, and symbolic artificial intelligence (AI), in this paper we present a real and fulfilled intelligent control architecture with application in bioprocess control.

Keywords: bioprocess, intelligent control, neural nets, fuzzy structure, hybrid techniques

Procedia PDF Downloads 424

Authors: P. Karthick, K. Mahesh

Abstract:

Video has become an increasingly significant component of our digital everyday contact. With the advancement of greater contents and shows of the resolution, its significant volume poses serious obstacles to the objective of receiving, distributing, compressing, and revealing video content of high quality. In this paper, we propose the primary beginning to complete a deep video compression model that jointly upgrades all video compression components. The video compression method involves splitting the video into frames, comparing the images using convolutional neural networks (CNN) to remove duplicates, repeating the single image instead of the duplicate images by recognizing and detecting minute changes using generative adversarial network (GAN) and recorded with long short-term memory (LSTM). Instead of the complete image, the small changes generated using GAN are substituted, which helps in frame level compression. Pixel wise comparison is performed using K-nearest neighbours (KNN) over the frame, clustered with K-means, and singular value decomposition (SVD) is applied for each and every frame in the video for all three color channels [Red, Green, Blue] to decrease the dimension of the utility matrix [R, G, B] by extracting its latent factors. Video frames are packed with parameters with the aid of a codec and converted to video format, and the results are compared with the original video. Repeated experiments on several videos with different sizes, duration, frames per second (FPS), and quality results demonstrate a significant resampling rate. On average, the result produced had approximately a 10% deviation in quality and more than 50% in size when compared with the original video.

Keywords: video compression, K-means clustering, convolutional neural network, generative adversarial network, singular value decomposition, pixel visualization, stochastic gradient descent, frame per second extraction, RGB channel extraction, self-detection and deciding system

Procedia PDF Downloads 188

Authors: Y. Zheng, M. Busi, A. F. Pedersen, M. A. Beltran, C. Gundlach

Abstract:

X-ray phase contrast imaging has shown to yield a better contrast compared to conventional attenuation X-ray imaging, especially for soft tissues in the medical imaging energy range. This can potentially lead to better diagnosis for patients. However, phase contrast imaging has mainly been performed using highly brilliant Synchrotron radiation, as it requires high coherence X-rays. Many research teams have demonstrated that it is also feasible using a laboratory source, bringing it one step closer to clinical use. Nevertheless, the requirement of fine gratings and high precision stepping motors when using a laboratory source prevents it from being widely used. Recently, a random phase object has been proposed as an analyzer. This method requires a much less robust experimental setup. However, previous studies were done using a particular X-ray source (liquid-metal jet micro-focus source) or high precision motors for stepping. We have been working on a much simpler setup with just small modification of a commercial bench-top micro-CT (computed tomography) scanner, by introducing a piece of sandpaper as the phase analyzer in front of the X-ray source. However, it needs a suitable algorithm for speckle tracking and 3D reconstructions. The precision and sensitivity of speckle tracking algorithm determine the resolution of the system, while the 3D reconstruction algorithm will affect the minimum number of projections required, thus limiting the temporal resolution. As phase contrast imaging methods usually require much longer exposure time than traditional absorption based X-ray imaging technologies, a dynamic phase contrast micro-CT with a high temporal resolution is particularly challenging. Different reconstruction methods, including neural network based techniques, will be evaluated in this project to increase the temporal resolution of the phase contrast micro-CT. A Monte Carlo ray tracing simulation (McXtrace) was used to generate a large dataset to train the neural network, in order to address the issue that neural networks require large amount of training data to get high-quality reconstructions.

Keywords: micro-ct, neural networks, reconstruction, speckle-based x-ray phase contrast

Procedia PDF Downloads 260

Authors: Tsu-Wang Shen, Shan-Chun Chang, Chih-Hsien Wang, Te-Chao Fang

Abstract:

For electrocardiogram (ECG) biometrics system, it is a tedious process to pre-install user’s high-intensity heart rate (HR) templates in ECG biometric systems. Based on only resting enrollment templates, it is a challenge to identify human by using ECG with the high-intensity HR caused from exercises and stress. This research provides a heartbeat segment method with slope-oriented neural networks against the ECG morphology changes due to high intensity HRs. The method has overall system accuracy at 97.73% which includes six levels of HR intensities. A cumulative match characteristic curve is also used to compare with other traditional ECG biometric methods.

Keywords: high-intensity heart rate, heart rate resistant, ECG human identification, decision based artificial neural network

Procedia PDF Downloads 436