在实际的应用中BP神经网络的初始权值和阈值的选择不是基于一个有效的选择

In practical applications, the selection of the initial weight and threshold of the BP neural network is not based on an effective selection mechanism, but has a large randomness, so it is difficult to select a global initial point. The overall goodness of these initial parameters is difficult to determine, and the initial values ​​are different, the convergence speed and performance of the network will be different. If the selection is not appropriate, the neural network training will not be smooth and easy to cause shocks. <br><br>Based on the above analysis, I modified this sentence to make it easier to understand.

In practical application, the selection of the initial weight and threshold of BP neural network is not based on an effective selection mechanism, but has a great randomness, so it is difficult to select a global initial point. The global excellence of these initial parameters is difficult to determine, the initial value is different, the convergence speed and performance of the network will be different, if not selected properly, will lead to the neural network training is not smooth, easy to cause shocks.<br><br>Based on the above analysis, I have modified this sentence to make it easier to understand.

In practical application, the selection of initial weights and thresholds of BP neural network is not based on an effective selection mechanism, but has great randomness, so it is difficult to select the global initial points. It is difficult to determine the global superiority of these initial parameters. If the initial values are different, the convergence speed and performance of the network will be different. If the initial parameters are not selected properly, the neural network training will not be smooth and easy to cause vibration.<br>Based on the above analysis, I modified this sentence to make it easier to understand.<br>