【摘要】 对人工神经网络的激励函数作了一些研究。用遗传算法优化各神经元Ｓｉｇｍｏｉｄ函数的温度参数来解决‘异或’问题，取得良好效果。在大量试验基础上总结了一些规律。表明了激励函数对于网络学习性能的重要作用。更多还原

【Abstract】 Researches on activation function of artificial neural network are conducted. The method uses the GA to optimize the thermo-parameters of each neuron. By using this method tosolve the ’XOR’ problem, better performance is achieved. Some conclusions are drawn through experiments. All of these indicate the importance of activation function to ANN’s learning ability.更多还原