目标问题的复杂程度和网络处理能力的适合程度是影响人工神经网络推广性的本质因素。为了衡量前向神经网络（ Ｆ Ｎ Ｎ）的处理能力，该文对 Ｆ Ｎ Ｎ 的插值误差进行了研究，得到了统计意义下 Ｆ Ｎ Ｎ 处理能力的估计值，进而定义了能间接反映神经网络推广性的推广性量度。该方法能够估计出适合目标问题的网络规模，应用于函数逼近和样本分类问题，仿真结果证实了该方法的有效性。 The complexity of the target problem and the suitability of network processing ability are the essential factors that affect the generalization of ANN. In order to measure the processing power of FNN, this paper studies the interpolation error of FNN, and obtains the estimated value of FNN processing capability under the statistical meaning, and then defines the neural network which can indirectly reflect the neural network Promote the promotion of measurement. This method can estimate the network size suitable for the target problem, apply to function approximation and sample classification, and the simulation results confirm the effectiveness of the proposed method.