Trade-Off Between Time Complexity and Accuracy of Perceptron Learning /sup */

In this paper, a modified perception leaning algorithm is proposed to reduce the computation time of learning. The proposed algorithm is easily programmed and can drastically decrease the time complexity of learning at the expense of only a little accuracy. Experimental results further show that this trade is worthy. Our proposed modified perceptron learning learning is then practical especially when noise exists in the training set or when the requirement of computational time is critical.