Smooth Group L1/2 Regularization for Pruning Convolutional Neural Networks

In this paper, a novel smooth group L1/2 (SGL1/2) regularization method is proposed for pruning hidden nodes of the fully connected layer in convolution neural networks. Usually, the selection of nodes and weights is based on experience, and the convolution filter is symmetric in the convolution neural network. The main contribution of SGL1/2 is to try to approximate the weights to 0 at the group level. Therefore, we will be able to prune the hidden node if the corresponding weights are all close to 0. Furthermore, the feasibility analysis of this new method is carried out under some reasonable assumptions due to the smooth function. The numerical results demonstrate the superiority of the SGL1/2 method with respect to sparsity, without damaging the classification performance.

Restricted Boltzmann Machines as Models of Interacting Variables

We study the type of distributions that restricted Boltzmann machines (RBMs) with different activation functions can express by investigating the effect of the activation function of the hidden nodes on the marginal distribution they impose on observed bi nary nodes. We report an exact expression for these marginals in the form of a model of interacting binary variables with the explicit form of the interactions depending on the hidden node activation function. We study the properties of these interactions in detail and evaluate how the accuracy with which the RBM approximates distributions over binary variables depends on the hidden node activation function and the number of hidden nodes. When the inferred RBM parameters are weak, an intuitive pattern is found for the expression of the interaction terms, which reduces substantially the differences across activation functions. We show that the weak parameter approximation is a good approximation for different RBMs trained on the MNIST data set. Interestingly, in these cases, the mapping reveals that the inferred models are essentially low order interaction models.

Performance Analysis of MAC Layer Protocols for WSN with Considering the Effects of Hidden Node Problem

The authors have requested that this preprint be withdrawn due to erroneous posting.

Classification of skin cancer images by applying simple evolving connectionist system

<span id="docs-internal-guid-eea5616b-7fff-5d26-eeb4-1d8c084ec93d"><span>Simple evolving connectionist system (SECoS) is one of data mining classification techniques that recognizing data based on the tested and the training data binding. Data recognition is achieved by aligning testing data to trained data pattern. SECoS uses a feedforward neural network but its hidden layer evolves so that each input layer does not perform epoch. SECoS distance has been modified with the normalized Euclidean distance formula to reduce error in training. This paper recognizes skin cancer by classifying benign malignant skin moles images using SECoS based on parameter combinations. The skin cancer classification has learning rate 1 of 0.3, learning rate 2 of 0.3, sensitivity threshold of 0.5, error threshold of 0.1 and MAPE is 0.5184845 with developing hidden node of 23. Skin cancer recognition by applying modified SECoS algorithm is proven more acceptable. Compared to other methods, SECoS is more robust to error variations.</span></span>

Performance Analysis of MAC Layer Protocols for WSN with Considering the Effects of Hidden Node Problem

Wireless sensor networks (WSNs) are consisting of a large number of sensor nodes that sense, gather, and process-specific data. Its importance is dedicated to its enormous application range. It could be used with industrial applications, agricultural applications, military applications, industrial applications, and a lot of other applications, which make it an open area for study by researchers and students. In this paper, the effects of the hidden node problem are studied on three different MAC protocols using various field distances and various numbers of nodes. This study provides the best number of nodes to be disseminated in a specific field distance depending on the needed performance metrics. Six performance metrics are used in this study, which is Goodput, Throughput, PDR, Residual Energy, Average Delay, and first and last node dead in the network. IEEE 802.11, IEEE 802.15.4, and TDMA protocols are the used protocols in this study. Eight different scenarios were proposed and implemented for this study. NS2 is used to construct the proposed scenarios. Results show that TDMA gives the best energy conservation and high delay time with high PDR, while IEEE 802.11 provides the best throughput and Goodput results and low delay time. A graphical view for the results was made to ease of study and analysis.