Denoising autoencoder

As explained in Data Without Labels

A denoising autoencoder is an autoencoder trained to reconstruct clean input data from corrupted or noisy input.

A denoising autoencoder reconstructs clean data from a corrupted version. For images, noise is added to the original image, the noisy result passes through an encoder-decoder network, and the output is compared with the original. section 9.7

The workflow has three essential elements:

• Create a noisy input by adding noise to an original image. section 9.7
• Pass the noisy input through the encoder-decoder architecture. section 9.7
• Compare the reconstructed output with the original image so the network learns a representation that removes noise. section 9.7

Image restoration

An original image is corrupted into a noisy image, which becomes the model input. The reconstructed output is evaluated against the original rather than against the corrupted input. section 9.7

Figure 9.5 The process of denoizing in an autoencoder. It starts with the original image; noise is added, which results in a noisy image, and then it is fed to the autoencoder.

Trace the denoising sequence from the original image through noise addition and autoencoder processing to the output. section 9.7

How Data Without Labels covers it

How deep it goes

Explains the corruption, encoder-decoder pass, and comparison target, but does not provide implementation details or mathematical derivations.

Examples

Uses original images, noise-corrupted images, and reconstructed outputs as the central visual example.

What it assumes

Assumes the basic autoencoder components of encoder, bottleneck, and decoder.

What you can do after

Apply the encoder-decoder process to reconstruct cleaner images and reduce nonlinear data dimensionality.

How it connects

Connects autoencoders with encoder-decoder architecture, latent representations, and nonlinear dimensionality reduction.

liveBookmanning.com

Read Data Without Labels on liveBook Get the book at manning.com

Also explained in

• 

  liveBookmanning.com
  Inside Deep Learning: Math, Algorithms, ModelsAn autoencoder trained to reconstruct clean data from corrupted input, used here as the analogy for sequence-to-sequence learning over sequences.

Compare how 2 books explain denoising autoencoder →

Related concepts

• autoencoderkind of · An autoencoder is a neural network that encodes input data into an internal representation and decodes it to reconstruct the input or an approximation.
• diffusion modelused by · A diffusion model is a generative model that learns to produce data by reversing a process that progressively adds noise to data.
• encoder-decoder architectureequivalent to · An encoder-decoder architecture is a model design in which an encoder transforms input into an intermediate representation and a decoder uses that representation to produce output.
• clean imageuses · A clean image is an uncorrupted image containing the original visual signal, used as a reference or reconstruction target.
• U-Netincludes · U-Net is an encoder-decoder neural-network architecture with downsampling and upsampling paths connected by skip connections for combining features at multiple spatial scales.

This explainer was generated by Manning's knowledge graph from the text of Data Without Labels. The section links point to the parts of the book it draws on.