Optical tensor core architecture for neural network training based on dual-layer waveguide topology and homodyne detection

We propose an optical tensor core(OTC) architecture for neural network training. The key computational components of the OTC are the arrayed optical dot-product units(DPUs). The homodyne-detection-based DPUs can conduct the essential computational work of neural network training, i.e., matrix-matrix multiplication. Dual-layer waveguide topology is adopted to feed data into these DPUs with ultra-low insertion loss and cross talk. Therefore, the OTC architecture allows a large-scale dot-product array and can be integrated into a photonic chip. The feasibility of the OTC and its effectiveness on neural network training are verified with numerical simulations.

Dimensionality Reduction with Input Training Neural Network and Its Application in Chemical Process Modelling

Many applications of principal component analysis (PCA) can be found in dimensionality reduction. But linear PCA method is not well suitable for nonlinear chemical processes. A new PCA method based on im-proved input training neural network (IT-NN) is proposed for the nonlinear system modelling in this paper. Mo-mentum factor and adaptive learning rate are introduced into learning algorithm to improve the training speed of IT-NN. Contrasting to the auto-associative neural network (ANN), IT-NN has less hidden layers and higher training speed. The effectiveness is illustrated through a comparison of IT-NN with linear PCA and ANN with experiments. Moreover, the IT-NN is combined with RBF neural network (RBF-NN) to model the yields of ethylene and propyl-ene in the naphtha pyrolysis system. From the illustrative example and practical application, IT-NN combined with RBF-NN is an effective method of nonlinear chemical process modelling.

Method to generate training samples for neural network used in target recognition

Training neural network to recognize targets needs a lot of samples.People usually get these samples in a non-systematic way,which can miss or overemphasize some target information.To improve this situation,a new method based on virtual model and invariant moments was proposed to generate training samples.The method was composed of the following steps：use computer and simulation software to build target object＇s virtual model and then simulate the environment,light condition,camera parameter,etc.;rotate the model by spin and nutation of inclination to get the image sequence by virtual camera;preprocess each image and transfer them into binary image;calculate the invariant moments for each image and get a vectors＇ sequence.The vectors＇ sequence which was proved to be complete became the training samples together with the target outputs.The simulated results showed that the proposed method could be used to recognize the real targets and improve the accuracy of target recognition effectively when the sampling interval was short enough and the circumstance simulation was close enough.

Bottom hole pressure estimation using hybridization neural networks and grey wolves optimization

An effective design and optimum production strategies of a well depend on the accurate prediction of its bottom hole pressure(BHP)which may be calculated or determined by several methods.However,it is not practical technically or economically to apply for a well test or to deploy a permanent pressure gauge in the bottom hole to predict the BHP.Consequently,several correlations and mechanistic models based on the knownsurfacemeasurementshave beendeveloped.Unfortunately,all these tools(correlations&mechanistic models)are limited to some conditions and intervals of application.Therefore,establish a global model that ensures a large coverage of conditions with a reduced cost and high accuracy becomes a necessity.In this study,we propose new models for estimating bottom hole pressure of vertical wells with multiphase flow.First,Artificial Neural Network(ANN)based on back propagation training(BP-ANN)with 12 neurons in its hidden layer is established using trial and error.The next methods correspond to optimized or evolved neural networks(optimize the weights and thresholds of the neural networks)with Grey Wolves Optimization(GWO),and then its accuracy to reach the global optima is compared with 2 other naturally inspired algorithms which are the most used in the optimization field:Genetic Algorithm(GA)and Particle Swarms Optimization(PSO).The models were developed and tested using 100 field data collected from Algerian fields and covering a wide range of variables.The obtained results demonstrate the superiority of the hybridization ANN-GWO compared with the 2 other hybridizations or with the BP learning alone.Furthermore,the evolved neural networks with these global optimization algorithms are strongly shown to be highly effective to improve the performance of the neural networks to estimate flowing BHP over existing approaches and correlations.

Computer-aided diabetic retinopathy diagnostic model using optimal thresholding merged with neural network

Purpose-Diabetic retinopathy(DR)is a central root of blindness all over the world.Though DR is tough to diagnose in starting stages,and the detection procedure might be time-consuming even for qualified experts.Nowadays,intelligent disease detection techniques are extremely acceptable for progress analysis and recognition of various diseases.Therefore,a computer-aided diagnosis scheme based on intelligent learning approaches is intended to propose for diagnosing DR effectively using a benchmark dataset.Design/methodology/approach-The proposed DR diagnostic procedure involves four main steps:(1)image pre-processing,(2)blood vessel segmentation,(3)feature extraction,and(4)classification.Initially,the retinal fundus image is taken for pre-processing with the help of Contrast Limited Adaptive Histogram Equalization(CLAHE)and average filter.In the next step,the blood vessel segmentation is carried out using a segmentation process with optimized gray-level thresholding.Once the blood vessels are extracted,feature extraction is done,using Local Binary Pattern(LBP),Texture Energy Measurement(TEM based on Laws of Texture Energy),and two entropy computations-Shanon’s entropy,and Kapur’s entropy.These collected features are subjected to a classifier called Neural Network(NN)with an optimized training algorithm.Both the gray-level thresholding and NN is enhanced by the Modified Levy Updated-Dragonfly Algorithm(MLU-DA),which operates to maximize the segmentation accuracy and to reduce the error difference between the predicted and actual outcomes of the NN.Finally,this classification error can correctly prove the efficiency of the proposed DR detection model.Findings-The overall accuracy of the proposed MLU-DA was 16.6%superior to conventional classifiers,and the precision of the developed MLU-DA was 22%better than LM-NN,16.6%better than PSO-NN,GWO-NN,and DA-NN.Finally,it is concluded that the implemented MLU-DA outperformed state-of-the-art algorithms in detecting DR.Originality/value-This paper adopts the latest optimization algorithm called MLU-DA-Neural Network with optimal gray-level thresholding for detecting diabetic retinopathy disease.This is the first work utilizes MLU-DA-based Neural Network for computer-aided Diabetic Retinopathy diagnosis.

How Deep Learning Networks could be Designed to Locate Mineral Deposits

Whether using a shallow neural network with one hidden layer,or a deep network with many hidden layers,the training data must represent subgroups of the deposit type being explored to be useful.Published examples of neural networks have mostly been limited to one individual mineral deposit for training.Variation of geologic features among deposits within a type are so large that a single deposit cannot provide proper information to train a neural net to generalize and guide exploration for other deposits.Models trained with only one deposit tend to be academic successes but are not of practical value in exploration for other deposits.This is why it takes much experience examining many deposits to properly train an economic geologist—a neural network is not any different.Two examples of shallow neural networks are used to demonstrate the power of neural networks to possibly locate undiscovered deposits and to provide some suggestions of how to deal with missing data.The training data needs to include information spatially related to known deposits and hopefully information from many different deposits of the type.Lessons learned from these and other examples point to a proposed sampling plan for data that could lead to a generalized neural network for exploration.In this plan,10 or more well-explored gold-rich porphyry copper deposits from around the world with 100 or more sample sites near and some distance from each deposit would probably capture important variability among such deposits and provide proper data to train and test a shallow neural network to predict locations of undiscovered deposits.