Gaussian process


In probability theory and statistics, a Gaussian process is a stochastic process, such that every finite collection of those random variables has a multivariate normal distribution, i.e. every finite linear combination of them is normally distributed. The distribution of a Gaussian process is the joint distribution of all those random variables, and as such, it is a distribution over functions with a continuous domain, e.g. time or space.
A machine-learning algorithm that involves a Gaussian process uses lazy learning and a measure of the similarity between points to predict the value for an unseen point from training data. The prediction is not just an estimate for that point, but also has uncertainty information—it is a one-dimensional Gaussian distribution.
For multi-output predictions, multivariate Gaussian processes
are used, for which the multivariate Gaussian distribution is the marginal distribution at each point.
For some kernel functions, matrix algebra can be used to calculate the predictions using the technique of kriging. When a parameterised kernel is used, optimisation software is typically used to fit a Gaussian process model.
The concept of Gaussian processes is named after Carl Friedrich Gauss because it is based on the notion of the Gaussian distribution. Gaussian processes can be seen as an infinite-dimensional generalization of multivariate normal distributions.
Gaussian processes are useful in statistical modelling, benefiting from properties inherited from the normal distribution. For example, if a random process is modelled as a Gaussian process, the distributions of various derived quantities can be obtained explicitly. Such quantities include the average value of the process over a range of times and the error in estimating the average using sample values at a small set of times.

Definition

A time continuous stochastic process is Gaussian if and only if for every finite set of indices in the index set
is a multivariate Gaussian random variable. That is the same as saying every linear combination of has a univariate normal distribution.
Using characteristic functions of random variables, the Gaussian property can be formulated as follows: is Gaussian if and only if, for every finite set of indices, there are real-valued, with such that the following equality holds for all
where denotes the imaginary unit such that.
The numbers and can be shown to be the covariances and means of the variables in the process.

Variance

The variance of a Gaussian process is finite at any time, formally

Stationarity

For general stochastic processes strict-sense stationarity implies wide-sense stationarity but not every wide-sense stationary stochastic process is strict-sense stationary. However, for a Gaussian stochastic process the two concepts are equivalent.
A Gaussian stochastic process is strict-sense stationary if, and only if, it is wide-sense stationary.

Example

There is an explicit representation for stationary Gaussian processes. A simple example of this representation is
where and are independent random variables with the standard normal distribution.

Covariance functions

A key fact of Gaussian processes is that they can be completely defined by their second-order statistics. Thus, if a Gaussian process is assumed to have mean zero, defining the covariance function completely defines the process' behaviour. Importantly the non-negative definiteness of this function enables its spectral decomposition using the Karhunen–Loève expansion. Basic aspects that can be defined through the covariance function are the process' stationarity, isotropy, smoothness and periodicity.
Stationarity refers to the process' behaviour regarding the separation of any two points and. If the process is stationary, it depends on their separation,, while if non-stationary it depends on the actual position of the points and. For example, the special case of an Ornstein–Uhlenbeck process, a Brownian motion process, is stationary.
If the process depends only on, the Euclidean distance between and, then the process is considered isotropic. A process that is concurrently stationary and isotropic is considered to be homogeneous; in practice these properties reflect the differences in the behaviour of the process given the location of the observer.
Ultimately Gaussian processes translate as taking priors on functions and the smoothness of these priors can be induced by the covariance function. If we expect that for "near-by" input points and their corresponding output points and to be "near-by" also, then the assumption of continuity is present. If we wish to allow for significant displacement then we might choose a rougher covariance function. Extreme examples of the behaviour is the Ornstein-Uhlenbeck covariance function and the squared exponential where the former is never differentiable and the latter infinitely differentiable.
Periodicity refers to inducing periodic patterns within the behaviour of the process. Formally, this is achieved by mapping the input to a two dimensional vector.

Usual covariance functions

There are a number of common covariance functions:
Here. The parameter is the characteristic length-scale of the process, is the Kronecker delta and the standard deviation of the noise fluctuations. Moreover, is the modified Bessel function of order and is the gamma function evaluated at. Importantly, a complicated covariance function can be defined as a linear combination of other simpler covariance functions in order to incorporate different insights about the data-set at hand.
Clearly, the inferential results are dependent on the values of the hyperparameters defining the model's behaviour. A popular choice for is to provide maximum a posteriori estimates of it with some chosen prior. If the prior is very near uniform, this is the same as maximizing the marginal likelihood of the process; the marginalization being done over the observed process values. This approach is also known as maximum likelihood II, evidence maximization, or empirical Bayes.

Continuity

For a Gaussian process, continuity in probability is equivalent to mean-square continuity,
and continuity with probability one is equivalent to sample continuity.
The latter implies, but is not implied by, continuity in probability.
Continuity in probability holds if and only if the mean and autocovariance are continuous functions. In contrast, sample continuity was challenging even for stationary Gaussian processes, and more challenging for more general processes.
As usual, by a sample continuous process one means a process that admits a sample continuous modification.

Stationary case

For a stationary Gaussian process some conditions on its spectrum are sufficient for sample continuity, but fail to be necessary. A necessary and sufficient condition, sometimes called Dudley-Fernique theorem, involves the function defined by
. Continuity of in probability is equivalent to continuity of at When convergence of to is too slow, sample continuity of may fail. Convergence of the following integrals matters:
these two integrals being equal according to integration by substitution The first integrand need not be bounded as thus the integral may converge or diverge. Taking for example for large that is, for small one obtains when and when
In these two cases the function is increasing on but generally it is not. Moreover, the condition
does not follow from continuity of and the evident relations and
Theorem 1. Let be continuous and satisfy Then the condition is necessary and sufficient for sample continuity of
Some history.
Sufficiency was announced by Xavier Fernique in 1964, but the first proof was published by Richard M. Dudley in 1967.
Necessity was proved by Michael B. Marcus and Lawrence Shepp in 1970.
There exist sample continuous processes such that they violate condition An example found by Marcus and Shepp is a random lacunary Fourier series
where are independent random variables with standard normal distribution; frequencies are a fast growing sequence; and coefficients satisfy The latter relation implies whence almost surely, which ensures uniform convergence of the Fourier series almost surely, and sample continuity of
Its autocovariation function
is nowhere monotone, as well as the corresponding function

Brownian motion as the integral of Gaussian processes

A Wiener process is the integral of a white noise generalized Gaussian process. It is not stationary, but it has stationary increments.
The Ornstein–Uhlenbeck process is a stationary Gaussian process.
The Brownian bridge is an example of a Gaussian process whose increments are not independent.
The fractional Brownian motion is a Gaussian process whose covariance function is a generalisation of that of the Wiener process.

Driscoll's zero-one law

Driscoll's zero-one law is a result characterizing the sample functions generated by a Gaussian process.
Let be a mean-zero Gaussian process with non-negative definite covariance function. Let be a Reproducing kernel Hilbert space with positive definite kernel.
Then
where and are the covariance matrices of all possible pairs of points, implies
What's more,
implies
This has significant implications when, as
As such, almost all sample paths of a mean-zero Gaussian process with positive definite kernel will lie outside of the Hilbert space.

Applications

A Gaussian process can be used as a prior probability distribution over functions in Bayesian inference. Given any set of N points in the desired domain of your functions, take a multivariate Gaussian whose covariance matrix parameter is the Gram matrix of your N points with some desired kernel, and sample from that Gaussian. For solution of the multi-output prediction problem, Gaussian process regression for vector-valued function was developed. In this method, a 'big' covariance is constructed, which describes the correlations between all the input and output variables taken in N points in the desired domain. This approach was elaborated in detail for the matrix-valued Gaussian processes and generalised to processes with 'heavier tails' like Student-t processes.
Inference of continuous values with a Gaussian process prior is known as Gaussian process regression, or kriging; extending Gaussian process regression to multiple target variables is known as cokriging. Gaussian processes are thus useful as a powerful non-linear multivariate interpolation tool. Gaussian process regression can be further extended to address learning tasks in both supervised and unsupervised learning frameworks.
Gaussian processes can also be used in the context of mixture of experts models, for example. The underlying rationale of such a learning framework consists in the assumption that a given mapping cannot be well captured by a single Gaussian process model. Instead, the observation space is divided into subsets, each of which is characterized by a different mapping function; each of these is learned via a different Gaussian process component in the postulated mixture.

Gaussian process prediction, or Kriging

When concerned with a general Gaussian process regression problem, it is assumed that for a Gaussian process observed at coordinates, the vector of values is just one sample from a multivariate Gaussian distribution of dimension equal to number of observed coordinates. Therefore, under the assumption of a zero-mean distribution,, where is the covariance matrix between all possible pairs for a given set of hyperparameters θ.
As such the log marginal likelihood is:
and maximizing this marginal likelihood towards θ provides the complete specification of the Gaussian process f. One can briefly note at this point that the first term corresponds to a penalty term for a model's failure to fit observed values and the second term to a penalty term that increases proportionally to a model's complexity. Having specified θ making predictions about unobserved values at coordinates x* is then only a matter of drawing samples from the predictive distribution where the posterior mean estimate A is defined as
and the posterior variance estimate B is defined as:
where is the covariance between the new coordinate of estimation x* and all other observed coordinates x for a given hyperparameter vector θ, and are defined as before and is the variance at point x* as dictated by θ. It is important to note that practically the posterior mean estimate is just a linear combination of the observations ; in a similar manner the variance of is actually independent of the observations. A known bottleneck in Gaussian process prediction is that the computational complexity of prediction is cubic in the number of points |x| and as such can become unfeasible for larger data sets. Works on sparse Gaussian processes, that usually are based on the idea of building a representative set for the given process f, try to circumvent this issue.

Bayesian neural networks as Gaussian processes

Bayesian neural networks are a particular type of Bayesian network that results from treating deep learning and artificial neural network models probabilistically, and assigning a prior distribution to their parameters. Computation in artificial neural networks is usually organized into sequential layers of artificial neurons. The number of neurons in a layer is called the layer width. As layer width grows large, many Bayesian neural networks reduce to a Gaussian process with a closed form compositional kernel. This Gaussian process is called the Neural Network Gaussian Process. It allows predictions from Bayesian neural networks to be more efficiently evaluated, and provides an analytic tool to understand deep learning models.

Software