[Study notes]
The above calculation shows that the neural network model of FIG really more in line with reality. First, a high probability of belonging to the luxurious buildings with swimming pool, followed by the price is relatively high. It does not belong to the low-end room with a swimming pool, and a lower price. Why is the same model, judged the result is not the same? 1) w3 relatively large, it is 0.8, w4 is relatively small, 0.001. Means that the expensive high-end facilities are very important for the judge. Mark -to-win @ Mark java community: For the low-end is almost useless. So find a good model of the process is to find them suitable parameters, this is called training model. House for judging this child. If we place a real estate broker, we actually in their daily work, constantly sum up to correct these parameters, continuous improvement of this artificial network model. Let's get this model for a wide range of input values are correct. Mark -to-win @ Mark java community: the discipline of artificial intelligence is doing this child, through continuous training, so that the parameters of your model more correct. So that for various values of the input, the determination result is correct.
Reprinted from the original article: https://blog.csdn.net/qq_44594249/article/details/100538434