# Sigmoid Function

A sigmoid function is any mathematical function whose graph has a characteristic S-shaped curve, commonly used as an activation function in artificial neural networks to map inputs to a range like (0,1).

A sigmoid function is any mathematical function whose graph has a characteristic S-shaped or sigmoid curve. In mathematics, a sigmoid function is bounded, differentiable, and defined for all real input values, with a positive derivative at each point. The term is most often associated with the logistic function, which maps any real number to a value between 0 and 1, but it also encompasses a broader family of curves used across statistics, engineering, and [machine-learning](https://www.wikiprompt.org/wiki/machine-learning).

In the context of [artificial-intelligence](https://www.wikiprompt.org/wiki/artificial-intelligence) and [neural-network](https://www.wikiprompt.org/wiki/neural-network) models, the term "sigmoid function" is frequently used as a synonym for the logistic function. This function played a foundational role in early [deep-learning](https://www.wikiprompt.org/wiki/deep-learning) architectures as an activation function, though it has been largely superseded by alternatives such as the rectified linear unit in many modern systems. Sigmoid curves also appear naturally as cumulative distribution functions in statistics, including those derived from the normal and Student's t distributions.

## Definition and Properties

A sigmoid function is a bounded, differentiable, real function defined for all real inputs, with a positive derivative at each point. Its graph is constrained by a pair of horizontal asymptotes as the input approaches positive or negative infinity. The function is monotonic, typically increasing, and its first derivative is bell-shaped. Conversely, the integral of any continuous, non-negative, bell-shaped function with a single local maximum yields a sigmoidal curve.

Sigmoid functions are convex for values below a particular inflection point and concave for values above it; in many common examples, this point is at zero. The logistic function, for instance, has its inflection point at x = 0, where its value is 0.5. The hyperbolic tangent function, another widely used sigmoid, ranges from -1 to 1 and is symmetric about the origin.

## Examples and Variants

Common sigmoid functions include the logistic function, the hyperbolic tangent (tanh), the arctangent, and the error function. The logistic function is defined as f(x) = 1 / (1 + e^(-x)), producing outputs in the open interval (0, 1). The hyperbolic tangent, tanh(x) = (e^x - e^(-x)) / (e^x + e^(-x)), maps inputs to (-1, 1) and is often preferred in [deep-learning](https://www.wikiprompt.org/wiki/deep-learning) because its outputs are zero-centered, which can improve convergence during training.

Special cases include the Gompertz curve, used in modeling systems that saturate at large values of x, and the ogee curve, applied in the spillway design of some dams. The Heaviside step function, which transitions instantaneously between 0 and 1, is a limiting case of sigmoid behavior. In statistics, sigmoid curves serve as cumulative distribution functions, such as the integrals of the logistic density, the normal density, and Student's t probability density functions.

## Role in Neural Networks

Sigmoid functions, particularly the logistic and hyperbolic tangent variants, have been used as activation functions for artificial neurons since the early days of [machine-learning](https://www.wikiprompt.org/wiki/machine-learning). An activation function introduces nonlinearity into a [neural-network](https://www.wikiprompt.org/wiki/neural-network), enabling it to learn complex patterns. The logistic sigmoid's output range of (0, 1) made it natural for probabilistic interpretations, such as predicting the probability of a binary outcome.

However, the logistic sigmoid suffers from the vanishing gradient problem: for large positive or negative inputs, its derivative approaches zero, which can slow or halt training in deep networks. This limitation motivated the development of alternatives like the rectified linear unit (ReLU) and its variants, which are now standard in most [transformer](https://www.wikiprompt.org/wiki/transformer)-based architectures, including those used in [large-language-model](https://www.wikiprompt.org/wiki/large-language-model) systems. Despite this shift, sigmoid functions remain important in specific contexts, such as the output layer of binary classifiers and in gating mechanisms within recurrent networks.

The logistic sigmoid is invertible, and its inverse is the logit function, which maps probabilities back to the real number line. This property underpins logistic regression, a fundamental statistical technique that bridges [machine-learning](https://www.wikiprompt.org/wiki/machine-learning) and classical statistics.

## Theoretical Classification

A mathematical theory, proposed by researcher Grebenc, classifies unitary sigmoid functions into three kinds based on their asymptotic behavior and oscillation near the asymptotes. A unitary sigmoid function is normalized to the unit range, typically with lower and upper asymptotes at 0 and 1, and can be expressed in the general form y = A * S(f(x)) + B, where S is an increasing sigmoid function, f(x) is a transformation of the independent variable, and A and B control scaling and translation.

Functions of the first kind approach their asymptotes monotonically without oscillation, including the logistic function and related generalizations. The second kind oscillates near the upper asymptote while preserving an overall sigmoid transition. The third kind oscillates near both the lower and upper asymptotes. The theory also lists 30 construction methods for generating sigmoid functions, ranging from algebraic transformations and integration of bell-shaped functions to solutions of differential equations, recursive schemes, and chaotic systems. These methods provide a systematic framework for designing sigmoid curves with tailored properties for specific applications in [artificial-intelligence](https://www.wikiprompt.org/wiki/artificial-intelligence) and other fields.

## Applications Beyond AI

Sigmoid functions extend beyond [neural-network](https://www.wikiprompt.org/wiki/neural-network) activation. In control theory and engineering, they model saturation effects in physical systems. In biology, the Gompertz curve describes growth processes such as tumor growth or population dynamics. In economics, sigmoid curves represent adoption of new technologies over time. The versatility of the S-shaped curve, with its smooth transition between two states, makes it a fundamental tool across disciplines, from [xerox-parc](https://www.wikiprompt.org/wiki/xerox-parc)-era computing research to modern [amazon-web-services](https://www.wikiprompt.org/wiki/amazon-web-services) cloud infrastructure that hosts [deep-learning](https://www.wikiprompt.org/wiki/deep-learning) workloads.

---
Source: https://www.wikiprompt.org/wiki/sigmoid
License: CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/)
Last updated: 2026-09-12T16:29:04.601455+00:00