Auto Encoder
Auto Encoder is an artificial neural network which is used to efficiently learn from the data using unsupervised learning. Auto Encoder reduces the dimensionality of the data. Is compresses the data, reconstructs it and again represents it in its original format. Auto Encoder is used to reduce the complexity by dimensionally reducing the weights. The main aim of auto Encoder is to increase the computational power. There are four parts of auto encoder.
Parts Of Auto Encoder:
- Encoder
- Bottleneck Layer
- Decoder
- Reconstruction Loss
1.Encoder: In this phase, we reduce the dimensionality of the data. It compresses the input data and converts it into the encoded format. The main aim of an encoder is to represent the data in encoded format.
2. Bottleneck layer: Bottleneck layer contains the compressed or encoded format of input data in it. In the bottleneck layer, we reduce the data to its smallest possible dimension. If the weight of the neural node decreases then its computation power increases and hence it will reduce the complexity.
3.Decoder: In this phase, we reconstruct the data from the encoded state to its original state. It focuses to reconstruct it closely to the input data set. Efficiency of decoder is calculated by its ability to convert it into proximate input data.
4.Reconstruction Loss: Reconstruction Loss measures the efficiency of the decoder. In this method, we are supposed to check how well did the decoder perform by identifying its closeness with the input data set. Reconstruction loss is also measured by the deviation of reconstructed data with the original data set
