[] Depth study of Formula confrontation Network (GAN) learned from the popular deepen understanding

Summary:

Formula confrontation Network (GAN) is a new area of ​​research, generate dialogue mainly used in the technical aspects of the image of image generation and natural language. Simply put: that is, the machine may need to generate new images and dialogue.

Popular understanding:

Let us assume a scenario: your neighbors are running a very cool party, so you want to go. But if you want to party, you need a special ticket, and this ticket has been sold out long ago.

For the organizers of the gathering, in order to be able to make the party successfully held, they hired a qualified security agencies. The main goal is to allow anyone to undermine the party. To do this, they placed a lot of guards in the entrance hall, to check the authenticity of tickets held by each person.

Considering you do not have any talent in the martial arts, and you especially want to go to the party, then the only solution is to use a very convincing fake tickets to deceive them. But the existence of this plan is a big bug-- you never really seen this in the end is what kind of tickets. So, in this case, if you only according to their own creativity design a ticket, then during the first attempt to want to fool guards almost impossible. In addition, unless you have a copy on the party ticket of a good, otherwise you'd better not put your face emerge.

To help solve the problem, you decide to call your friend Bob to do the job for you. Bob's task is very simple. He will try to use your fake passport to enter the party. If he is rejected, he will return and tell you something about the real ticket should be what kind of advice.

Based on this feedback, you can make a new version of the ticket, and then handed over to Bob, go to the ticket office to try. This process is repeated until you are able to design a perfect ticket "copies."

Deepen understanding:

Reference suggestions:

https://www.baidu.com/link?url=X7oteFvad-yIpDSGx9TnETCUFYZ_4uk6tptqExwqdr6RNBqLbdCd5nvt18loQRkZnlqKnemVFezgzepYn9UWq5AOgnjWndaq4ZhA_hJanxe&wd=&eqid=8b601db60001528a000000065ce7fed3

Guess you like

Origin blog.csdn.net/weixin_39589455/article/details/90524074