Autoencoder
An autoencoder is a neural network that encodes input data into an internal representation and decodes it to reconstruct the input or an approximation.
Inside Deep Learning: Math, Algorithms, Models
A neural network trained to reconstruct its input, typically by compressing it into a lower-dimensional latent representation and decoding it back.
- How it is used
- Uses input reconstruction for label-free representation learning and reconstruction error for outlier detection.
- How deep it goes
- Develops encoder–decoder notation, bottleneck and denoising constraints, and PyTorch implementations through PCA and deeper fully connected models.
- Examples
- Uses 28 × 28 image reconstructions, noisy-input comparisons, and exercises on outlier detection and k-means clustering.
- What you can do after
- Build it from scratch in PyTorch with encoder and decoder modules, MSE training, and optional bottleneck or training-only noise.
Read how Inside Deep Learning: Math, Algorithms, Models explains autoencoder →
Build AI Drug Discovery Pipelines
An autoencoder is a neural model that encodes inputs into a latent representation and decodes that representation to reconstruct the inputs.
- How it is used
- Builds a compressed molecular representation for exploring chemical space and designing molecules.
- How deep it goes
- Develops tokenization, embeddings, encoder-decoder architecture, reconstruction loss, and benchmark evaluation before extending the model to a variational autoencoder.
- Examples
- Uses SMILES reconstruction and the MOSES molecular generation benchmark to illustrate and evaluate the model.
- What you can do after
- Build it from scratch with PyTorch for molecular reconstruction and evaluate reconstruction, latent-space quality, and generation.
Read how Build AI Drug Discovery Pipelines explains autoencoder →
Outlier Detection in Python
An autoencoder is a neural network trained to encode an input into a representation and reconstruct the original input.
- How it is used
- Uses reconstruction error to estimate outlierness in tabular and image data.
- How deep it goes
- Explains encoder, bottleneck, decoder, and reconstruction, then demonstrates library detectors rather than deriving their training algorithms.
- Examples
- Demonstrates a PyOD detector on encoded KDD Cup records and an alibi-detect detector on clean and noise-perturbed CIFAR10 images.
- What you can do after
- Apply it via PyOD or alibi-detect to score candidate outliers and inspect image reconstructions.
Math and Architectures of Deep Learning
A neural network model trained to compress inputs into a latent representation and reconstruct the original inputs from that representation.
- How it is used
- Learns compact input representations by reconstructing inputs without human-provided labels.
- How deep it goes
- Develops encoder-decoder training, reconstruction losses, convolutional implementations, and nonlinear comparison with PCA; defers autoencoder training details during transpose-convolution discussion.
- Examples
- Uses 32 × 32 image encoders and decoders, normalized image reconstruction, transpose-convolution upsampling, and MNIST reconstruction comparisons.
- What you can do after
- Build an image autoencoder in PyTorch with convolutional encoding, transposed-convolution decoding, and reconstruction loss.
Read how Math and Architectures of Deep Learning explains autoencoder →
Data Without Labels
An autoencoder is a neural-network model that learns to encode data into a lower-dimensional representation and reconstruct the original input from it.
- How it is used
- Uses autoencoders for unsupervised feature learning, dimensionality reduction, reconstruction-based anomaly detection, and denoising.
- How deep it goes
- Explains encoder–bottleneck–decoder training choices and implements a basic model, while leaving the mathematics of variants to further reading.
- Examples
- Builds a Keras autoencoder for MNIST digits and plots test images beside their reconstructions.
- What you can do after
- Apply it via Keras to train a basic digit autoencoder and inspect reconstructions on test data.
Evolutionary Deep Learning: Genetic algorithms and neural networks
An autoencoder is a neural network that learns to encode inputs into a representation and decode that representation to reconstruct the inputs.
- How it is used
- Uses reconstruction without labels to learn feature encodings for images and to optimize AE architectures evolutionarily.
- How deep it goes
- Works through a convolutional Keras build, training, genetic encoding, mutation, crossover, and validation-loss evaluation; variational AEs are only summarized.
- Examples
- Uses Fashion-MNIST training images, MNIST latent-space visualization with t-SNE, and evolved convolutional architectures.
- What you can do after
- Build it from scratch as a convolutional Keras autoencoder and evolve balanced encoder-decoder architectures with DEAP.
Read how Evolutionary Deep Learning: Genetic algorithms and neural networks explains autoencoder →
Learn Generative AI with PyTorch
A neural network that learns to compress inputs into a latent representation and reconstruct them, used here for digit-image generation and reconstruction.
- How it is used
- Introduces learned image reconstruction as a foundation for variational autoencoders.
- How deep it goes
- Develops the encoder, decoder, training loop, and reconstruction inspection, while leaving validation-based parameter tuning aside.
- Examples
- Uses grayscale MNIST digits, 20-value encodings, and side-by-side original and reconstructed images.
- What you can do after
- Build an autoencoder from scratch to reconstruct handwritten digits and inspect its results.
Read how Learn Generative AI with PyTorch explains autoencoder →
Graph Neural Networks in Action
An autoencoder is a neural network that encodes inputs into latent representations and decodes those representations to reconstruct the inputs.
- How it is used
- Uses reconstruction to learn compact representations and, for graphs, predict missing links.
- How deep it goes
- Explains the encoder–latent space–decoder pipeline and outlines graph training steps without deriving the loss mathematically.
- Examples
- Illustrates image reconstruction and graph link prediction on the Amazon Products dataset.
- What you can do after
- Build it from scratch as a graph autoencoder by defining components, preparing graph data, training with edge data, and testing predictions.
Read how Graph Neural Networks in Action explains autoencoder →
Delivering Machine Learning Projects: From design to deployment
An autoencoder is a neural network that learns to encode inputs into a lower-dimensional representation and reconstruct the original inputs.
- How it is used
- Uses reconstruction error to detect anomalous news articles during regional feature development.
- How deep it goes
- Explains the operational use of reconstruction errors but does not detail architecture, training procedure, or tuning.
- Examples
- Uses historical regional news articles and per-article reconstruction errors as the concrete detection example.
- What you can do after
- Evaluate anomalous articles by measuring their reconstruction errors against a learned regional baseline.
Read how Delivering Machine Learning Projects: From design to deployment explains autoencoder →
Transfer Learning for Natural Language Processing
A model that learns to reconstruct its input, typically by encoding it into a latent representation and decoding that representation.
- How it is used
- Uses reconstruction to project target-domain samples into a source-domain feature space for sentiment domain adaptation.
- How deep it goes
- Explains reconstruction and latent dimensionality, then specifies a shallow network, its training objective, and classifier evaluation.
- Examples
- Uses IMDB movie reviews and book and DVD reviews from the Multi-Domain Sentiment Dataset.
- What you can do after
- Applies it via Keras to reconstruct vectors and evaluate projected inputs with a sentiment classifier.
Read how Transfer Learning for Natural Language Processing explains autoencoder →
Financial AI in Practice
An autoencoder is a neural network trained to reconstruct its input, typically by encoding it into a lower-dimensional representation and decoding it.
- How it is used
- Finds potentially novel fraud by flagging deviations from learned normal transaction patterns.
- How deep it goes
- Explains normal-only data preparation, network architecture, training, validation, reconstruction scoring, and operational threshold considerations.
- Examples
- Uses credit card transactions and a 30-to-14-to-30 architecture to illustrate normal-data training and reconstruction.
- What you can do after
- Apply it via Keras with a TensorFlow backend to train a normal-transaction autoencoder and evaluate reconstruction scores.
Deep Learning with R, Third Edition
An autoencoder is a neural-network model that learns to encode inputs into representations and decode those representations back into approximations of the inputs.
- How it is used
- Compresses inputs into latent representations and reconstructs them, with denoising as its strongest practical application.
- How deep it goes
- Explains encoder–decoder reconstruction, latent constraints, denoising, and probabilistic extensions without deriving their mathematics.
- Examples
- Uses image reconstruction, low-quality JPEG denoising, smartphone super-resolution, and repeated denoising from pure noise.
- What you can do after
- Evaluate/understand how encoder–decoder models reconstruct images and why denoising and super-resolution can invent plausible detail.
Read how Deep Learning with R, Third Edition explains autoencoder →
Deep Learning with Python, Third Edition
A neural network trained to reconstruct its input, often by compressing and then decoding it, used here for denoising images.
- How it is used
- Uses reconstruction to learn image representations, denoise degraded inputs, and support latent-space image generation.
- How deep it goes
- Explains encoder-decoder reconstruction, sparse codes, variational extensions, and iterative denoising without detailing implementation.
- Examples
- Uses reconstructed images, noisy JPEGs, low-resolution photographs, hallucinated moon craters, and pure-noise-to-image generation.
- What you can do after
- Evaluate/understand when reconstruction, denoising, latent representations, or diffusion-based generation fit an image task.
Read how Deep Learning with Python, Third Edition explains autoencoder →
Introduction to Generative AI, Second Edition
An autoencoder is a neural network that learns to encode data into a compressed representation and reconstruct the original data from it.
- How it is used
- Explains image and video creation or alteration within synthetic media, especially deepfake face-swapping.
- How deep it goes
- Defines encoding, latent representations, and reconstruction, then sketches face-swapping; it stops short of implementation details.
- Examples
- Uses facial expressions, face angles, skin tones, and a single-encoder, two-decoder face swap.
- What you can do after
- Evaluate/understand how compression and reconstruction support face-swapping and why synthetic-media trust is difficult.
Read how Introduction to Generative AI, Second Edition explains autoencoder →
Machines that Think
An autoencoder is a neural-network architecture that encodes an input into a constrained latent representation and decodes that representation to reconstruct the input.
- How it is used
- Uses reconstruction through a constrained latent bottleneck to learn data structure rather than labels.
- How deep it goes
- Defines the encoder, latent space, and decoder, then links reconstruction to feature learning and applications without detailing training procedures.
- Examples
- Uses chair descriptions, faces, digits, corrupted images or audio, unusual patterns, and dimensionality reduction as illustrations.
- What you can do after
- Understand how reconstruction-based compression can support denoising, anomaly detection, and dimensionality reduction.
Also covered in
- Data-Driven Cybersecurity
- Machine Learning Algorithms in Depth
- Build a Text-to-Image Generator (from Scratch)
- Sutskever's List
- Causal AI
What to know first
- mean squared errorMean squared error is the arithmetic mean of the squared differences between corresponding predicted and reference numerical values.
- deep learningDeep learning is a machine-learning approach that uses neural networks with multiple computational layers to learn representations from data.
- generative modelingGenerative modeling is a machine-learning approach that learns patterns or a data distribution to generate new samples resembling the data.
- language modelA language model is a computational model that represents patterns in linguistic sequences to assign probabilities, predict, generate, or evaluate language.
Related concepts
- encoderhas part · An encoder is a component or process that transforms input into an encoded or internal representation for subsequent processing, transmission, or decoding.
- decoderhas part · A decoder is a component that transforms an encoded representation into values, reconstructed data, or generated output.
- variational autoencoderincludes · A variational autoencoder is a generative neural network that learns a probabilistic latent representation of data and decodes samples from it to reconstruct or generate data.
- deep learningkind of · Deep learning is a machine-learning approach that uses neural networks with multiple computational layers to learn representations from data.
- neural networkkind of · A neural network is a computational model composed of interconnected, parameterized units or transformations that learn to map inputs to outputs.
- Kerasuses · Keras is an open-source, high-level deep-learning API for building, training, and evaluating neural-network models.
Questions about autoencoder
What is autoencoder?
An autoencoder is a neural network that encodes input data into an internal representation and decodes it to reconstruct the input or an approximation.
Which Manning books explain autoencoder?
Inside Deep Learning: Math, Algorithms, Models, Build AI Drug Discovery Pipelines, Outlier Detection in Python, Math and Architectures of Deep Learning, Data Without Labels, Evolutionary Deep Learning: Genetic algorithms and neural networks, Learn Generative AI with PyTorch, Graph Neural Networks in Action and 7 more.
What should I know before learning autoencoder?
Manning books that teach autoencoder usually introduce mean squared error, deep learning, generative modeling and language model first.
What kinds of autoencoder are there?
Kinds of autoencoder include variational autoencoder.
What is autoencoder a kind of?
Autoencoder is a kind of deep learning. Deep learning is a machine-learning approach that uses neural networks with multiple computational layers to learn representations from data.
What are the parts of autoencoder?
Autoencoder includes encoder and decoder.
These guides are generated from the text of Manning books by Manning's knowledge graph and link to the sections they draw on. The full books are available on liveBook and at manning.com.














