LSTM

RNNs are multilayer neural networks that are used to recognize patterns in sequences of data. By sequences of data, we mean text, handwriting, numerical times series (coming for example from sensors), log entries, and so on. The algorithms involved here have a temporal dimension too: they take time (and this is the main difference with CNNs) and sequence both into account. For a better understanding of the need for RNNs, we have to look at the basics of feedforward networks first. Similar to RNNs, these networks channel information through a series of mathematical operations performed at the nodes of the network, but they feed information straight through, never touching a given node twice. The network is fed with input examples that are then transformed into an output: in simple words, they map raw data to categories. Training happens for them on labeled inputs, until the...