➢ A Recurrent Neural Network(RNN) is a class of Artificial Neural Network in which,the
neurons are able to send feedback signal to each other. ➢ RNN has a "Memory", which remember all the information about what has been calculate. ➢ ANN/CNN accepts a fixed-sized vectors as an input and produce a fixed-sized vectors as an output. ➢ RNN allows us to operate over vector sequence in input,output or in both. What is Vanishing Gradient Problem? ➢ When we do back-propagation moving backward in network and calculating gradient of loss(Errors) with respect to weight,the gradient tends to get smaller and smaller as we keep on moving backward in the network. ➢ That means the neurons in earlier layers learn very slowly as compared to the neurons that are in later layers,the earlier layers in network are slowest to train. ➢ So, the process of reducing gradient as we go backward in network is called Vanishing Gradient. Solution of Vanishing Gradient: ➢ We do not use sigmoid and tanh as an activation function which causes vanishing gradient probelms. ➢ Instead of use this we use ReLu as an activation function to overcome this problems. What is Long Short-Term Memory (LSTM)? ➢ Long short-term memory (LSTM) is a recurrent neural network (RNN) that selectively remembers patterns for long durations of time. ➢ A LSTM unit is composed of a cell and three gates which control the flow of information. 1.Forget gate 2.Input gate 3.Output gate Forgot gate: ➢ Forget gate will decide what to forget from the previous memory you need. Input gate: ➢ The input gate will decide what to accept inside the neuron,update gate will update the memory. Output gate: ➢ The output gate will give the output as long term memory. what is Gated Recurrent unit(GRU)? ➢ They are two types of gates in GRU 1.Update Gate 2.Reset Gate Update Gate: ➢ The update gate access forgot gate an input get inside the LSD update get besides what information to throw away and what informtion to add in neuron. Reset Gate: ➢ Reset gate is used to decide How much pass information to forgot. Why we need RNN? ➢ Because of their internal memory, RNN’s can remember important things about the input they received, which allows them to be very precise in predicting what’s coming next. ➢ This is why they're the preferred algorithm for sequential data like time series, speech, text, financial data,audio, video, weather and much more. ➢ Recurrent neural networks can form a much deeper understanding of a sequence and its context compared to other algorithms. when do you need to use a RNN? ➢ Whenever there is a sequence of data and that temporal dynamics that connects the data is more important than the spatial content of each individual frame. ➢ Since RNNs are being used in the software behind Siri and Google Translate, recurrent neural networks show up a lot in everyday life.