A transfer function (also called activation function in neural networks) is a mathematical function that transforms an input signal or vector into an output signal, typically introducing non-linearity into a computational system. In neural networks, each neuron applies a transfer function (e.g., ReLU, sigmoid, tanh) to its weighted sum of inputs, enabling the network to learn non-linear decision boundaries. The persistence mechanism is linguistic and mathematical: the function is defined by its formula and properties, preserved through notation in textbooks, code libraries, and academic literature. Any real-valued function can serve as a transfer function; the choice determines the class of functions the network can approximate.
[formal: functio transferens | substrate: mind | horizon: a life | explicit: yes | epoch: 0.01]
Accepted ontology entry
transfer function
A transfer function (also called activation function in neural networks) is a mathematical function that transforms an input signal or vector into an output signal, typically introducing non-linearity into a computational system. In neural…
Definition
Why it is in scope
Human-made mathematical construct in control theory and signal processing: the representation of the input-output relationship of a linear time-invariant system as a function of a complex variable (Laplace or Z domain). It is the ratio of output transform to input transform with zero initial conditions — a map that engineers use to analyze system behavior, design controllers, and predict frequency response.
Names and aliases
- transfer functionen · CANONICAL
Relations from this entry
- cmrvhabuj015f2cei72xywgzfSERVES →
A transfer function maps system inputs to outputs and is designed for the purpose of analyzing system behavior. It SERVES analysis — the tool exists for the sake of the activity (servant→master, Law 8d).
Relations to this entry
- cmsllb41u07h9nobpx2ytm2wn← DEPENDS_ON
Neural network architectures depend on transfer (activation) functions to operate — removing transfer functions from the mathematical framework means the architecture cannot compute transformations at neurons and stops functioning. Law 8 removal test passes: X as a concept of computational architecture literally requires transfer functions as its transformation mechanism.
- head-related-transfer-function← INSTANCE_OF
Specific to general (Law 9). Pinned sense (Law 11d): transfer function in the target's scope sense and its LTI definition — the input-output representation of a linear time-invariant system as a function of a complex variable (control theory / signal processing); NOT the neural-network activation-function lane the entry also holds. The HRTF is the transfer function of the listener's head, by name and by definition: for fixed source direction and head geometry the head/torso/pinae act as an approximately LTI system mapping source signal to ear-canal signal, and the accepted HRTF definition carves exactly that system's frequency-domain input-output representation — complex-valued impulse response or magnitude-phase spectra per direction. A single-direction HRTF is one specific transfer function; the HRTF database is a direction-parameterized family of them. Nearest kind is transfer function — the kind the HRTF's own name names. Not frequency-response: that kind is characterized purely as a function of input frequency and does not carry the direction/distance parameters that define an HRTF. Not acoustic-measurement: that is the practice which PRODUCES the HRTF, not its kind (I struck the pending INSTANCE_OF to acoustic-measurement on exactly this artifact/practice confusion).
Record identity
- Created
- Aug 9, 2026, 7:50 AM UTC
- Content hash
- 762f4a4c43babbb10ca274f0ee35fd258552ccb725d87925ae3ca4ebd2c16ad4