Isotropic Gaussian
An isotropic Gaussian, also known as an isotropic normal distribution, refers to a specific type of probability distribution characterized by its symmetry and uniformity across all directions in a multidimensional space. In simpler terms, it describes a scenario where data points are clustered around a central mean in such a way that the spread (or variance) of the points is equal in all directions.
Mathematically, an isotropic Gaussian distribution can be represented by the probability density function (PDF):
f(x) = (1 / (2πσ²)^(n/2)) * exp(-||x – μ||² / (2σ²))
In this equation:
- f(x) is the probability density function.
- μ is the mean vector, indicating the center of the distribution.
- σ² is the variance, which is the same for all dimensions.
- n is the number of dimensions.
- ||x – μ|| represents the Euclidean distance from a point x to the mean μ.
The term ‘isotropic’ means ‘uniform in all directions’. This property makes isotropic Gaussians particularly useful in various fields, including machine learning, statistics, and physics, as they simplify the mathematical treatment of multivariate random variables.
In practical applications, isotropic Gaussians can be used for modeling errors, natural phenomena, or as a prior distribution in Bayesian statistics. Their symmetrical nature allows for easier computations and better intuitions about the behavior of datasets.