Recurrent Neural Networks (RNN) are used for sequential modeling on datasets where high autocorrelation exists among observations. For example, predicting patient journeys using their historical dataset or predicting the next words in given sentences. The main commonality among these problem statements is that input length is not constant and there is a sequential dependence. Standard neural network and deep learning models are constrained by fixed size input and produce a fixed length output. For example, deep learning neural networks built on occupancy datasets have six input features and a binomial outcome.
Generative models in machine learning domains are referred to as models that have an ability to generate observable data values. For example, training a generative model on an images repository to generate new images like it. All generative models aim to compute the joint distribution over given datasets, either implicitly or...