Deep Learning Essentials

Generative Adversarial Networks

Recently, one popular field of machine learning that has seen the use of deep learning techniques is generative learning. Generative learning can be defined as a technique for learning joint probability estimates, P(x,y) from features and labels. It builds a probabilistic model of labels and can be robust to missing data and noisy data. Additionally, such models can also be used to generate samples, which can be further used to train advanced machine learning models. A discriminative model on the other hand learns a function that maps data x with label y, thereby learning a conditional probability distribution of P(y|x). Though in recent years, discriminative models have shown promising results in machine ...