An Improved Genetic Algorithm for Automated Convolutional Neural Network Design

Extracting the features from an image is a cumbersome task. Initially, this task was performed by domain experts through process known as handcrafted feature design. A deep embedding technique convolutional neural networks (CNNs) later solved problem introducing learning concept, which CNN directly provided with images. This then learns of image, are subsequently given input to further layers for intended like classification. CNNs have demonstrated astonishing performance in several practicable applications last few years. Nevertheless, pursuance primarily depends upon their architecture, expertise and type investigated problem. On other hand, researchers who do not proficiency using CNNs, it has been very difficult explore topic statements. In paper, we come up rank gradient descent-based optimized genetic algorithm automatically find architecture design that vigorously competent exploring best maneuvering tasks proposed algorithm, there no requirement pre- post-processing, implies fully mechanized. The validation on conventional benchmarked datasets done comparing run time graphics processing unit (GPU) throughout training assessing accuracy various measures. experimental results show accomplishes better more persistent ‘classification accuracy’ than original CIFAR fifty percent less intensive computing resources individual entire population.