Gaussian process

As explained in AI-Powered Search

A Gaussian process is a collection of random variables whose finite subsets have joint Gaussian distributions and that defines a probability distribution over functions.

A Gaussian process is a statistical model that predicts values while providing a probability distribution describing prediction certainty. In search, it helps balance exploiting a ranking model’s known strengths with exploring poorly understood query–document combinations. section 12.3.2 section 12.3

Think of exploration as surveying a river with only scattered depth measurements. Nearby dates are easier to predict and safer to exploit; distant, unobserved dates carry more uncertainty and may reveal useful knowledge, but also introduce risk. A Gaussian process estimates both expected outcomes and uncertainty so exploration can target promising gaps. section 12.3.2

A search exploration pipeline can use a Gaussian process as follows:

• Train GaussianProcessRegressor on logged feature combinations and their observed relevance grades. section 12.3.2
• Generate candidate combinations of ranking features, such as name matching, description matching, and promotion status. section 12.3.2
• Predict each candidate’s grade and standard deviation, retaining both expected performance and uncertainty. section 12.3.2
• Calculate expected improvement from opportunity, probability of improvement, and predictive uncertainty, then rank candidates for exploration. section 12.3.2

Selecting an unexplored product

For the query "transformers dvd", logged judgments train the model on available ranking signals. The exploration score favors a feature vector with useful expected improvement, and the search engine retrieves a product matching that vector. One selected result is "Transformers: The Complete Series [25th Anniversary ...]" with UPC 826663114164. section 12.3.2

Figure 12.5 Exploring a river, with uncertainty of the river’s depth increasing as we get further away from past observations. How would we pick a time of year that is both safe and maximally increases our knowledge of the river’s depth?

Use the river-depth figure to compare safe exploitation near known observations with riskier exploration where predictive uncertainty grows. section 12.3.2

How AI-Powered Search covers it

How deep it goes

Explains probabilistic prediction, uncertainty, expected improvement, and implementation, but stops short of deriving Gaussian-process mathematics.

Examples

Uses river-depth observations, feature combinations, and the “transformers dvd” product-selection example.

What it assumes

Assumes logged judgments, ranking features, and relevance grades while building intuition from basic observations and uncertainty.

What you can do after

Apply it via scikit-learn to rank search candidates by expected improvement and select products for exploration.

How it connects

Connects Gaussian processes to active learning, learning to rank, click models, presentation bias, and expected improvement.

liveBookmanning.com

Read AI-Powered Search on liveBook Get the book at manning.com

Also explained in

• 

  liveBookmanning.com
  Bayesian Optimization in ActionA Gaussian process is a probabilistic model over functions that defines predictive distributions, including means and uncertainties, at unobserved inputs.
• 

  liveBookmanning.com
  Machine Learning Algorithms in DepthA Gaussian process is a collection of random variables such that every finite subset has a jointly Gaussian distribution, defining a probability distribution over functions.

Compare how 3 books explain Gaussian process →

Related concepts

• language modelkind of · A language model is a computational model that represents patterns in linguistic tokens to estimate sequence probabilities and predict or generate language from context.
• mean functionhas part · A mean function maps an input to the expected value of a random function before conditioning on observed data.
• covariance functionhas part · A covariance function, also called a kernel, specifies the covariance between function values at pairs of inputs in a Gaussian process.
• Bayesian optimizationused by · Bayesian optimization is a sequential method for optimizing an expensive or unknown objective by using a probabilistic model to guide evaluations.
• kernel functionuses · A kernel function is a mathematical function that measures similarity between pairs of inputs, often representing an inner product in an implicit feature space.
• radial basis function kerneluses · A radial basis function kernel is a similarity function whose value depends on the distance between inputs and typically decreases as that distance increases.

This explainer was generated by Manning's knowledge graph from the text of AI-Powered Search. The section links point to the parts of the book it draws on.