Data Science Journal

Because simple feed forward network assumes that data points are independent and treats each data point individually while sequential data is characterized by dependency, that is each data point is some how dependent on previous data point. Hence feed forward network fails to learn dependency information and consequently results in poor predictions. Here comes RNN for rescue yay.

Because it provides  extensive built-in support for deep learning mathematical functions for neural networks auto differentiation (for back propagation) and optimizers Source