Table 2.

Procedure of SFGAN.

The Procedure of SFGAN
1. Generate $n$ noise samples $z=\{z^1,\cdots ,z^n\}.$ Each $z^i$ is vector given by $z^i=(z_1^i,z_2^i,\dots ,z_d^i)$, where d is the number of dimensions, and each component of $z^i$ is randomly generated according to a uniform distribution in the interval [−1,1] with pdf $p_z.$
2. The generator $G$ generates fake data $G(z)$ based on the generated samples $z.$
3. $n$ samples $x=\{x^1,\cdots ,x^n\}$ are taken from real data with pdf $p_{data}.$
4. Auto-encoder $D$ receives $G(z)$ and $x$ as inputs and produces $D\left(G(z)\right)$ and $D(x)$ as outputs.