# Creation of MASTER

The creation of MASTER essentially involves constructing a new set of DNNs from scratch and then training it. The training is based on the principle of GANs. In other words, we need two DNNs: “AIRTIST-DNN,” which corresponds to the art generation and “Discriminator,” that improves the quality of the creation process.

The two networks are connected as shown in figure 1 below, where the lower-left corner is “AIRTIST-DNN,” and the middle is the “Discriminator” network. At the beginning of training, the weight parameters of both networks are random values. The AIRTIST-DNN network is a generative model that can map a set of random numbers to an artwork.

However, the generated works may not have any artistic value at all initially. At the same time, the Discriminator network is also trained, using pre-selected artwork samples and artworks generated by AIRTIST-DNN as sample data. The training goal of the two networks is to minimise the difference between the created artwork and the real artwork.

More strictly, the overall objective is to make the statistical distribution of AIRTIST-DNN generated data approximate as much as possible the distribution of the input artwork samples. During the training process, the weight parameters of the two networks are continuously adjusted. This training process can be regarded as the interaction of an artist and a critic in a zero-sum game.

After reaching the Nash equilibrium, the generated artwork will bear a high resemblance to the target master.

![Figure 1: Creation of the MASTER](/files/Ix99oQM6MU4VY5dBmzpw)


---

# Agent Instructions: Querying This Documentation

If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question.

Perform an HTTP GET request on the current page URL with the `ask` query parameter:

```
GET https://airtist.gitbook.io/product-docs/airtist-ecosystem/inception/creation-of-master.md?ask=<question>
```

The question should be specific, self-contained, and written in natural language.
The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.