A novel MapReduce-based deep convolutional neural network algorithm

Deep convolutional neural networks (DCNNs), with their complex network structure and powerful feature learning and feature expression capabilities, have been remarkable successes in many large-scale recognition tasks. However, with the expectation of memory overhead and response time, along with the increasing scale of data, DCNN faces three non-rival challenges in a big data environment: excessive network parameters, slow convergence, and inefficient parallelism. To tackle these three problems, this paper develops a deep convolutional neural networks optimization algorithm (PDCNNO) in the MapReduce framework. The proposed method first pruned the network to obtain a compressed network in order to effectively reduce redundant parameters. Next, a conjugate gradient method based on modified secant equation (CGMSE) is developed in the Map phase to further accelerate the convergence of the network. Finally, a load balancing strategy based on regulate load rate (LBRLA) is proposed in the Reduce phase to quickly achieve equal grouping of data and thus improving the parallel performance of the system. We compared the PDCNNO algorithm with other algorithms on three datasets, including SVHN, EMNIST Digits, and ISLVRC2012. The experimental results show that our algorithm not only reduces the space and time overhead of network training but also obtains a well-performing speed-up ratio in a big data environment.

Differential Evolution for Neural Networks Optimization

In this paper, a Neural Networks optimizer based on Self-adaptive Differential Evolution is presented. This optimizer applies mutation and crossover operators in a new way, taking into account the structure of the network according to a per layer strategy. Moreover, a new crossover called interm is proposed, and a new self-adaptive version of DE called MAB-ShaDE is suggested to reduce the number of parameters. The framework has been tested on some well-known classification problems and a comparative study on the various combinations of self-adaptive methods, mutation, and crossover operators available in literature is performed. Experimental results show that DENN reaches good performances in terms of accuracy, better than or at least comparable with those obtained by backpropagation.

Modern trends in the mathematical simulation of pressure-driven membrane processes

The presented article is an attempt to evaluate the progress in the development of the mathematical simulation of the pressure-driven membrane processes. It was considered more than 170 articles devoted to the simulation of reverse osmosis, nanofiltration, ultrafiltration, and microfiltration and the others published between 2000 and 2010 years. Besides the conventional approaches, which include the irreversible thermodynamics, diffusion and pore flow (and models which consider the membrane surface charge for nanofiltration process), the application of the methods the computational fluid dynamics, artificial neural networks, optimization, and economic analysis have been considered. The main trends in this field have been pointed out, and the areas of using approaches under consideration have been determined. The technological problems which have been solved using the mentioned approaches have also been considered. Although the question of the concentration polarization has not been considered separately, it was defined that, in many cases, the sufficiently accurate model cannot be designed without considering this phenomenon. The findings allow evaluating more thoroughly the development of the simulation of pressure-driven membrane processes. Moreover, the review allows choosing the strategy of the simulation of the considered processes. Keywords: membrane, simulation, model, reverse osmosis nanofiltration, ultrafiltration, microfiltration.

Optimized artificial neural network for classification of biological data

Biological data suffers from the problem of high dimensionality which makes the process of multi-class classification difficult and also these data have elements that are incomplete and redundant. Breast Cancer is currently one of the most pre-dominant causes of death in women around the globe. The current methods for classifying a tumour as malignant or benign involve physical procedures. This often leads to mental stress. Research has now sought to implement soft computing techniques in order to classify tumours based on the data available. In this paper, a novel classifier model is implemented using Artificial Neural Networks. Optimization is done in this neural network by using a meta-heuristic algorithm called the Whale Swarm Algorithm in order to make the classifier model accurate. Experimental results show that new technique outperforms other existing models.