Performance Improvement of Competitive Associative Nets via Statistical Learning Schemes and Its Engineering Applications

这项研究通过统计学习方案提高了竞争联想网络CAN2的性能，并在工程应用中显示了其有效性。这里的CAN2是一个人工神经网络学习有效的分段线性逼近的非线性函数。我们已经（1）构建和分析了CAN 2的统计学习算法，（2）应用于非线性时变对象的控制，（3）应用于语音处理，（4）应用于距离图像处理。在许多实际应用中，改进的方法具有灵活性，鲁棒性和有效性。

This research study has improved the performance of competitive associative nets called CAN2 by means of statistical learning schemes and shown its effectiveness in engineering applications. Here theCAN2 is an artificial neural net for learning efficient piecewise linear approximation of nonlinear function. We have(1) constructed and analyzed statistical learning algorithms for the CAN2,(2) applied to control of nonlinear time-varying plants,(3) applied to speech processing, and(4) applied to range image processing. The improved methods are shown to have flexibility, robustness, and effectiveness in many practical applications.