V-Shaped Backpropagation Algorithm for Image Classification

With the growth of artificial intelligence, machine learning becomes a major requirement of modern computer systems and artificially intelligent devices. For this purpose, many researchers developed different techniques to train a device to recognize and classify different objects. The Neural network is one of these techniques. In a neural network, if we fix its structure and apply it on different datasets, then it may be a possibility that a smaller structure is sufficient for a particular dataset. A fixed sized structure which may increase computation time for learning. To overcome this problem, we design a V-shaped model. The proposed V-shaped approach can reshape hidden layers according to inputs. Size of all hidden layers of the proposed V-shaped model completely depends on inputs. Size of inputs is a pivot for a total of hidden layers. The proposed V-shaped model performs better in terms of recognition accuracy and computational time.