Using neural network capacity evaluation, I expect output is according to the index system of empowerment, the actual output value of the training after the needs and expectations do contrast, equivalent to or according to the weight of ah, then using the neural network is a no meaning?

CodePudding user response:

According to the weight calculation score this itself is the most simple neural network (full connection layer), you just use a network to another network learning...

This issue really want to become you think meaningful, so first you need to have a premise, according to the weight calculation of score is not reasonable, need to use a new method to evaluate, you need to the following steps:
1. Data (one for each sample to ability score, score even can be simplified to good and bad two)
2. Training network, even if the input is only two, good and bad output can also be a score
So the network with you the meaning of human,

CodePudding user response:

reference 1st floor Zhu Mingde response:

according to weighting score this itself is the most simple neural network (full connection layer), you just use a network to another network learning...

This issue really want to become you think meaningful, so first you need to have a premise, according to the weight calculation of score is not reasonable, need to use a new method to evaluate, you need to the following steps:
1. Data (one for each sample to ability score, score even can be simplified to good and bad two)
2. Training network, even if the input is only two, good and bad output can also be a score
So the network has the meaning of what you people,

Thank you for your reply, want to ask this kind of good input differential similar levels of input values, can output a score of neural network model by which a neural network to realize?? A fuzzy neural network??

CodePudding user response:

Might as well assume that a good score is 1, to 0.5, the difference of 0,
Arbitrary neural network, finally one output for one dimensional full connection layer, then sigmoid constraint to the (0 ~ 1), the result is a score of 0 to 1

CodePudding user response:

reference Zhu Mingde reply: 3/f

presupposes a good score is 1, to 0.5, the difference of 0,
Any neural network, finally one output for one dimensional full connection layer, then sigmoid constraint to the (0 ~ 1), the result is a 0 ~ 1 score

Understand, thank you for your advice, but also want to ask once, at the time of training, such as I have 10 grade evaluation indicators, each indicator to enter good score is 1, to 0.5, the difference of 0; So how do you determine the expected output score

CodePudding user response:

In not, want to ask you some questions about this post