Recurrent Neural Network
Definition of Recurrent Neural Network
Recurrent neural networks can be seen as finding patterns in time series data and using them to predict the future. The neuron input requires not only the data at the current time point, but also the output results at the previous time point. The neuron returns the output results to the input end, and so on. In the time dimension, it is expanded as shown below:
At present, recurrent neural networks have been widely used in natural language processing, mainly speech recognition and language models. Among them, handwriting recognition is the earliest research result that successfully used RNN.
Characteristics of Recurrent Neural Networks
Recurrent Neural Network (RNN) has the concept of directed loop, which can be used to deal with the problem of the connection between inputs. The structure of the directed loop is shown in the figure below:
Compared with the feedforward neural network FNN, the recurrent neural network is more consistent with the structure of biological neural network.
Applications of Recurrent Neural Networks
In the field of natural language processing, such as word vector expression, sentence validity check, word tagging, etc., recurrent neural networks are successful, among which the most widely used model is the LSTM (Long Short-Term Memory) model. Currently, recurrent neural networks are mainly used in the following aspects:
- Language Modeling and Generating Text
- Machine Translation
- Speech Recognition
- Generating Image Descriptions
Reference Links:
1.https://zh.wikipedia.org/wiki/Recurrent Neural Networks