A three-stage learning algorithm for deep multilayer perceptron (DMLP) with effective weight initialisation based on sparse auto-encoder is proposed in this paper, which aims to overcome difficulties in training deep neural networks with limited training data in high-dimensional feature space. At the first stage, unsupervised learning is adopted using sparse auto-encoder to obtain the initial weights of the feature extraction layers of the DMLP. At the second stage, error back-propagation is used to train the DMLP by fixing the weights obtained at the first stage for its feature extraction layers. At the third stage, all the weights of the DMLP obtained at the second stage are refined by error back-propagation. Network structures and values of learning parameters are determined through cross-validation, and test datasets unseen in the cross-validation are used to evaluate the performance of the DMLP trained using the three-stage learning algorithm. Experimental results show that the proposed method is effective in combating overfitting in training deep neural networks.
Different ANN architectures of MLP have been trained by BP and used to analyze Landsat TM images. Two different approaches have been applied for training: an ordinary approach (for one hidden layer M-H1-L & two hidden layers M-H1-H2-L) and one-against-all strategy (for one hidden layer (M-H1-1)xL, & two hidden layers (M-H1-H2-1)xL). Classification accuracy up to 90% has been achieved using one-against-all strategy with two hidden layers architecture. The performance of one-against-all approach is slightly better than the ordinary approach
Semantic segmentation is effective in numerous object classification tasks such as autonomous vehicles and scene understanding. With the advent in the deep learning domain, lots of efforts are seen in applying deep learning algorithms for semantic segmentation. Most of the algorithms gain the required accuracy while compromising on their storage and computational requirements. The work showcases the implementation of Convolutional Neural Network (CNN) using Discrete Cosine Transform (DCT), where DCT exhibit exceptional energy compaction properties. The proposed Adaptive Weight Wiener Filter (AWWF) rearranges the DCT coefficients by truncating the high frequency coefficients. AWWF-DCT model reinstate the convolutional l
... Show MoreIn current generation of technology, a robust security system is required based on biometric trait such as human gait, which is a smooth biometric feature to understand humans via their taking walks pattern. In this paper, a person is recognized based on his gait's style that is captured from a video motion previously recorded with a digital camera. The video package is handled via more than one phase after splitting it into a successive image (called frames), which are passes through a preprocessing step earlier than classification procedure operation. The pre-processing steps encompass converting each image into a gray image, cast off all undesirable components and ridding it from noise, discover differen
... Show MoreInformation from 54 Magnetic Resonance Imaging (MRI) brain tumor images (27 benign and 27 malignant) were collected and subjected to multilayer perceptron artificial neural network available on the well know software of IBM SPSS 17 (Statistical Package for the Social Sciences). After many attempts, automatic architecture was decided to be adopted in this research work. Thirteen shape and statistical characteristics of images were considered. The neural network revealed an 89.1 % of correct classification for the training sample and 100 % of correct classification for the test sample. The normalized importance of the considered characteristics showed that kurtosis accounted for 100 % which means that this variable has a substantial effect
... Show MoreThe transition of customers from one telecom operator to another has a direct impact on the company's growth and revenue. Traditional classification algorithms fail to predict churn effectively. This research introduces a deep learning model for predicting customers planning to leave to another operator. The model works on a high-dimensional large-scale data set. The performance of the model was measured against other classification algorithms, such as Gaussian NB, Random Forrest, and Decision Tree in predicting churn. The evaluation was performed based on accuracy, precision, recall, F-measure, Area Under Curve (AUC), and Receiver Operating Characteristic (ROC) Curve. The proposed deep learning model performs better than othe
... Show MoreThis paper uses Artificial Intelligence (AI) based algorithm analysis to classify breast cancer Deoxyribonucleic (DNA). Main idea is to focus on application of machine and deep learning techniques. Furthermore, a genetic algorithm is used to diagnose gene expression to reduce the number of misclassified cancers. After patients' genetic data are entered, processing operations that require filling the missing values using different techniques are used. The best data for the classification process are chosen by combining each technique using the genetic algorithm and comparing them in terms of accuracy.
Recognizing facial expressions and emotions is a basic skill that is learned at an early age and it is important for human social interaction. Facial expressions are one of the most powerful natural and immediate means that humans use to express their feelings and intentions. Therefore, automatic emotion recognition based on facial expressions become an interesting area in research, which had been introduced and applied in many areas such as security, safety health, and human machine interface (HMI). Facial expression recognition transition from controlled environmental conditions and their improvement and succession of recent deep learning approaches from different areas made facial expression representation mostly based on u
... Show More