GAN is a generative adversarial network . In a generative model uses unsupervised learning approaches .in generative model there are samples in data that is the input variable but it lacs the output variable y and uses only input variable to train the generative model and recognizes patterns from the input variable that is unknown and based on the training data only.
In supervised learning , we are aligned to predictive models this is called as discriminative model ,here the model has to discriminate as to which class the example belongs to ,generative models are able to generate new examples which are not only similar to examples but are distinguishable as well. GAN are deep learning model that is used for unsupervised learning ,
where two computing neural network compete with each other .
Most of the Gans now a days use deep convolutional general adversarial network.
Gans consist of two sub models they are generative model and discriminator model. Generative network takes a sample and generates a sample of data Discriminator network decides whether the data is generated or taken from the real sample using a binary classification problem with the help of sigmoid which gives the output in the form 0 or 1.
Train the discriminator and freeze the generator ,which means the training set of the generator is turned as false and the network will only do the forward and no back propagation will be applied
Train the generator and then freeze the discriminator , in this we get the result from the first phase and can use them to make better from the previous state to try and fool the discriminator better.
Challenges faced :
- Problem of stability between generator and discriminator.
- Problem to determine the position of the object.
- Problem in understanding the perspective.
- Problem in understanding the global object.
- Prediction of next frame in a video .
- Text to image generation.
- Image to image generation.