Building robust neural networks using different loss functions

The paper considers the problem of building robust neural networks using different robust loss functions. Applying such neural networks is reasonably when working with noisy data, and it can serve as an alternative to data preprocessing and to making neural network architecture more complex. In order to work adequately, the error back-propagation algorithm requires a loss function to be continuously or two-times differentiable. According to this requirement, two five robust loss functions were chosen (Andrews, Welsch, Huber, Ramsey and Fair). Using the above-mentioned functions in the error back-propagation algorithm instead of the quadratic one allows obtaining an entirely new class of neural networks. For investigating the properties of the built networks a number of computational experiments were carried out. Different values of outliers’ fraction and various numbers of epochs were considered. The first step included adjusting the obtained neural networks, which lead to choosing such values of internal loss function parameters that resulted in achieving the highest accuracy of a neural network. To determine the ranges of parameter values, a preliminary study was pursued. The results of the first stage allowed giving recommendations on choosing the best parameter values for each of the loss functions under study. The second stage dealt with comparing the investigated robust networks with each other and with the classical one. The analysis of the results shows that using the robust technique leads to a significant increase in neural network accuracy and in a learning rate.

Neural network-based hardware classifier using CORDIC algorithm

This paper designs a novel classification hardware framework based on neural network (NN). It utilizes COordinate Rotation DIgital Computer (CORDIC) algorithm to implement the activation function of NNs. The training was performed through software using an error back-propagation algorithm (EBPA) implemented in C++, then the final weights were loaded to the implemented hardware framework to perform classification. The hardware framework is developed in Xilinx 9.2i environment using VHDL as programming languages. Classification tests are performed on benchmark datasets obtained from UCI machine learning data repository. The results are compared with competitive classification approaches by considering the same datasets. Extensive analysis reveals that the proposed hardware framework provides more efficient results as compared to the existing classifiers.

Enhancement of Shadow Region in an Image using Artificial Neural Network

Image captured in darker region increases complexities in processing and extracting vital information. Enhancement of such images helps us to retrieve important data and various tools are available for the same. Proposed system uses multi layer feed forward artificial neural network. Error Back Propagation algorithm is used in training process. Desired data is obtained using log transformation method. The proposed model is trained to enhance only shadow part of an image. The results shows enhancement in the darker region and is expected to improve more by changing different parameters in the above methodology.

A Pruning Neural Network Model in Credit Classification Analysis

Nowadays, credit classification models are widely applied because they can help financial decision-makers to handle credit classification issues. Among them, artificial neural networks (ANNs) have been widely accepted as the convincing methods in the credit industry. In this paper, we propose a pruning neural network (PNN) and apply it to solve credit classification problem by adopting the well-known Australian and Japanese credit datasets. The model is inspired by synaptic nonlinearity of a dendritic tree in a biological neural model. And it is trained by an error back-propagation algorithm. The model is capable of realizing a neuronal pruning function by removing the superfluous synapses and useless dendrites and forms a tidy dendritic morphology at the end of learning. Furthermore, we utilize logic circuits (LCs) to simulate the dendritic structures successfully which makes PNN be implemented on the hardware effectively. The statistical results of our experiments have verified that PNN obtains superior performance in comparison with other classical algorithms in terms of accuracy and computational efficiency.

Flower image classification with basket of features and multi layered artificial neural networks

Artificial intelligence is penetrating most of the classification and recognition tasks performed by a computer. This work proposes to classify flower images based on features extracted during segmentation and after segmentation using multiple layered neural networks. The segmentation models used are watershed, wavelet, wavelet fusion model, supervised active contours based on shape, color and Local binary pattern textures and color, fused textures based active contours. Multi-dimension feature vectors are constructed from these segmented results for each indexed flower image labelled with their name. Each feature becomes input to a neuron in various feature layers and error back propagation algorithm with convex optimization structure trains these multiple feature layers. Testing with different flower images sets from multiple sources resulted in average classification accuracy of 92% for shape, color and texture supervised active contour segmented flower images.