Deep learning

As explained in Outlier Detection in Python

Deep learning is machine learning based on neural networks with multiple computational layers that learn hierarchical representations.

Deep learning uses neural networks with more than a few layers. Each layer calculates numeric values from its inputs and passes them to the next; the final layer produces a prediction. For an image, inputs can be pixel values; for a table, they can be feature values. In outlier detection, deep networks provide several ways to judge whether an object is unusual. section 16.1 section 16.0

Figure 16.1 Example of a neural network. The square points at the left are the input. The circular points are neurons. During prediction, as input is given to the neural network and neurons calculate their values (floating point numbers each neuron will contain) one layer at a time (left to right) based on the input, with the final (rightmost) layer containing the prediction for the input.

Follow the input from left to right: neurons calculate values one layer at a time, and the rightmost layer holds the prediction. section 16.1

For images, deep-learning approaches include:

• Reconstruct the image after compressing it, then use reconstruction error to estimate how unusual it is. section 16.3.1
• Use a generative model to assess unusual data, including approaches that attempt to regenerate it. section 16.3.1
• Inspect a classifier’s confidence: low probabilities across its known classes can suggest an out-of-distribution image. section 16.3.1
• Create embeddings and apply detectors such as Isolation Forest or k-nearest neighbors; embeddings made for classification or regression may not work well for this purpose. section 16.3.1
• Alter an image, perhaps by rotating it, and test whether a model identifies the alteration confidently; unusual images may yield more errors or lower confidence. section 16.3.1

Deep learning is particularly useful for image, video, and audio outlier detection, but it is typically less useful for tabular data than established alternatives. Deep networks can require substantial training data, run more slowly, and be harder to tune. Choose an approach according to the data and the signal it provides, rather than treating any one score as a universal test. section 16.0 section 16.3.1

How Outlier Detection in Python covers it

How deep it goes

Explains layer-by-layer prediction and surveys image-detection approaches, but does not specify training procedures or implementations in the supplied evidence.

Examples

Uses image classification, low-confidence bird classifications, and rotated images to illustrate predictions and potential outlier signals.

What it assumes

Assumes basic statistics and some Python familiarity while introducing neural-network inputs, layers, and predictions without requiring prior outlier-detection knowledge.

What you can do after

Evaluate/understand it by comparing image-outlier signals and weighing deep learning’s data, speed, tuning, and tabular-data trade-offs.

How it connects

Connects neural networks with reconstruction error, generative models, classifier confidence, embeddings, Isolation Forest, and k-nearest neighbors.

liveBookmanning.com

Read Outlier Detection in Python on liveBook Get the book at manning.com

Also explained in

• 

  liveBookmanning.com
  Machine Learning for Tabular Data: XGBoost, Deep Learning, and AIA machine learning approach based on neural networks with multiple learned layers of representation.
• 

  liveBookmanning.com
  Inside AIA deep neural network is a neural network containing multiple hidden layers that learn hierarchical representations from simpler features.
• 

  liveBookmanning.com
  Data Without LabelsA deep learning model is a machine-learning model composed of multiple layers of learned representations.
• 

  liveBookmanning.com
  Designing Deep Learning Systems: A guide for software engineersA machine learning model that uses multiple layers of learned representations to make predictions or decisions.
• 

  liveBookmanning.com
  Evolutionary Deep Learning: Genetic algorithms and neural networksA deep learning model is a neural network with multiple learned layers that learns hierarchical representations from data.
• 

  liveBookmanning.com
  Effective Data AnalysisDeep learning is a machine-learning approach that uses multilayered neural networks to learn representations and patterns from data.
• 

  liveBookmanning.com
  WordPress Power Toolkit: Harness AI to build next-level websitesDeep learning is a machine-learning approach that uses multilayer artificial neural networks to learn hierarchical representations from large amounts of data.
• 

  liveBookmanning.com
  Delivering Machine Learning Projects: From design to deploymentA deep neural network is a neural-network model composed of multiple computational layers that learn hierarchical representations from data.
• 

  liveBookmanning.com
  Data-Driven CybersecurityDeep learning is a machine-learning approach that uses neural networks with multiple learned layers to represent complex patterns.
• 

  liveBookmanning.com
  Introduction to Generative AI, Second EditionDeep learning is a machine-learning subfield based on neural networks with multiple hidden layers.

Compare how 11 books explain deep learning →

Related concepts

• machine learningkind of · Machine learning is a field of computational methods in which systems learn patterns or relationships from data to make predictions, decisions, or other outputs.
• language modelincludes · A language model is a computational model that represents patterns in linguistic tokens to estimate sequence probabilities and predict or generate language from context.
• autoencoderincludes · An autoencoder is a neural network that encodes input data into an internal representation and decodes it to reconstruct the input or an approximation.
• neural networkuses · A neural network is a computational model of interconnected, parameterized units that learns transformations from inputs to outputs.
• GPT modelincludes · A GPT model is a generative pretrained Transformer language model that produces text autoregressively by predicting successive tokens from preceding context.
• recurrent neural networkincludes · A recurrent neural network is a neural network that processes sequences by updating an internal state across successive inputs, allowing earlier inputs to influence later computations.

This explainer was generated by Manning's knowledge graph from the text of Outlier Detection in Python. The section links point to the parts of the book it draws on.