人工神经网络BP算法的改进和结构的自调整

本文解决了BP神经网络结构参数和学习速率的选取问题,并对传统的BP算法进行了改进,提出了BP神经网络动态全参数自调整学习算法,又将其编制成计算机程序,使得隐层节点和学习速率的选取全部动态实现,减少了人为因素的干预,改善了学习速率和网络的适应能力。计算结果表明：BP神经网络动态全参数自调整算法较传统的方法优越。训练后的神经网络模型不仅能准确地拟合训练值,而且能较精确地预测未来趋势。     

Simultaneous electrochemical and piezoelectric microgravimetry measurements of C60 films with a quartz crystal microbalance in a dimethylformamide-water mixed solvent solution containing potassium ions

Simultaneous electrochemistry and piezoelectric microgravimetry studies using electrochemical quartz crystal microbalance were carried out on solution cast C₆₀ films. These films could be reduced in dimethylformamide + water mixture containing potassium ions. It was established that the first reduction was a completely surface-confined process, while the second reduction produced dissolved species. Restricting the potential range only to the first reduction, we could observe reversible electrochemical transformation of the films. Measuring the mass changes of the electrodes during the electrochemical processes, we found that the simple voltammograms (single reduction and oxidation peaks) actually cover simultaneous processes. Both the reduction and the oxidation involved the movements of anions, cations and solvent molecules in and out of the films. It was measured that the initial behavior of the films varied randomly, probably depending on the unpredictable structure of the film produced by solution casting. However, the films adjusted themselves to the experimental conditions displaying steady-state behavior after a certain number of potential scans. This steady-state behavior could be reached independently of the original structure, or the amount of C₆₀ put on the surface, but it could be perturbed easily by altering the conditions. The films were self-adjusted to the measurements.Dedicated to Professor György Horányi on the occasion of his 70th birthday in recognition of his outstanding contributions to electrochemistry