Recurrent Neural Network and its Various Architecture Types
- April 2, 2018
- Posted by: RSIS
- Category: Computer Science and Engineering
International Journal of Research and Scientific Innovation (IJRSI) | Volume V, Issue III, March 2018 | ISSN 2321–2705
Recurrent Neural Network and its Various Architecture Types
Trupti Katte
Assistant Professor, Computer Engineering Department, Army Institute of Technology, Pune, Maharashtra, India
Abstract-Recurrent neural network are network with dynamic capabilities to generate and process temporal information. Recurrent neural network are network can deep learn the input with its various architecture and identify outputs. LSTM network model was the first RNN with greatest achievement in pattern recognition contest in 2014. RNN can be used in its various architecture forms depending on the needs. Here we provide brief summary of various RNN networks up to now with Why? How? When? to use this network. Different architecture showed that recurrent neural network is mostly network with feed -forward and if want to store some information fed-back
Keywords-Recurrent neural Network, RNN, LSTM, NTM, Neural.
I. INTRODUCTION
A Neuron/ Nerve cell is a cell that carries electrical impulses. Neurons are most important part in brain and are the basic units of nervous system. Neural Network is the computer system which modeled based on concept of human brain and nervous systems neuron. Artificial Neural Network are the computing systems which are based on concept of biological neural networks. It consists of connections and units/Node (input, output, hidden).Nodes represent artificial neuron and connections represent direction of flow of information from one node to other.
The Recurrent Neural Network (RNN) is the network with loops, which allows information to persist in network. RNN has feed-back connection to the network itself, which allows activations to flow back in a loop, learn sequences and information to persist. RNN are extremely powerful in modeling sequential data, speech or text and applied on non-sequential data to train in a non-sequential manner. RNN can be used for image, video captioning, word prediction, word translation, image processing, speech recognition, speech processing [1], natural language processing, music processing applications, etc.
