To enhance fault tolerance ability of the feedforward neural networks (NNs for short) implemented in hardware, we discuss the learning algorithm that converges without adding extra neurons and a large amount of extra learning time and cycles. Our algorithm modified from the standard backpropagation algorithm (SBPA for short) limits synaptic weights of neurons in range during learning phase. The upper and lower bounds of the weights are calculated according to the average and standard deviation of them. Then our algorithm reupdates any weight beyond the calculated range to the upper or lower bound. Since the above enables us to decrease the standard deviation of the weights, it is useful in enhancing fault tolerance. We apply NNs trained with other algorithms and our one to a character recognition problem. It is shown that our one is superior to other ones in reliability, extra learning time and/or extra learning cycles. Besides we clarify that our algorithm never degrades the generalization ability of NNs although it coerces the weights within the calculated range.

In this paper we propose a learning algorithm to enhance the fault tolerance of feedforward neural networks (NNs for short) by manipulating the gradient of sigmoid activation function of the neuron. We assume stuck-at-0 and stuck-at-1 faults of the connection link. For the output layer, we employ the function with the relatively gentle gradient to enhance its fault tolerance. For enhancing the fault tolerance of hidden layer, we steepen the gradient of function after convergence. The experimental results for a character recognition problem show that our NN is superior in fault tolerance, learning cycles and learning time to other NNs trained with the algorithms employing fault injection, forcible weight limit and the calculation of relevance of each weight to the output error. Besides the gradient manipulation incorporated in our algorithm never spoils the generalization ability.

A block-matching-based self-organizing map (BMSOM) is presented. Finding a winner is carried out for each block, which is a set of neurons arranged in square. The proposed learning process updates the reference vectors of all of the neurons in a winner block. Then, the degrees of vector modifications are mainly controlled by the size (i.e., the number of neurons) of the winner block. To prevent a single cluster with neurons from splitting into some disjointed clusters, the restriction of the block size is imposed in the beginning of learning. At the main stage, this restriction is canceled. In BMSOM learning, the size of a winner block does not always decrease monotonically. The formula used to update the reference vectors is basically uncontrolled by time. Therefore, even if a map is in a nonstationary environment, training the map is probably pursued without interruption to adjust time-controlled parameters such as learning rate. Numerical results demonstrate that the BMSOM makes it possible to improve the plasticity of maps in a nonstationary environment and incremental learning.

In this paper we propose a MIN (Multistage Interconnection Network) whose performance in the faulty case degrades as gracefully as possible. We focus on a two-dilated baseline network as a sort of MIN. The link connection pattern in our MIN is determined so that all the available paths established between an input terminal and an output terminal via an identical input of a SE (Switching Element) in some stage will never pass through an identical SE in the next stage. Extra links are useful in improving the performance of the MIN and do not complicate the routing scheme. There is no difference between our MIN and others constructed from a baseline network with regard to numbers of links and cross points in all SEs. The theoretical computation and simulation-based study show that our MIN is superior to others in performance, especially in robustness against concentrated SE faults in an identical stage.

Data detection based on self organizing maps is presented for hematopoietic tumor patients. Learning data for the maps are generated from the screening data of examinees. The incomplete screening data without some item values is then supplemented by substituting averaged non-missing item values. In addition, redundant items, which are common to all the data and tend to have an unfavorable influence on data detection, are eliminated by a genetic algorithm and/or an immune algorithm. It is basically judged, by observing the label of a winner neuron in the map, whether the data presented to the map belongs to the class of hematopoietic tumors. Some experimental results are provided to show that the proposed methods achieve the high probability of correctly identifying examinees as hematopoietic tumor patients.