diskriminator

Diskriminator is a term used in several fields to denote an entity that distinguishes between categories or classes, or a device that converts a qualitative distinction into a measurable signal. The spelling with a k appears in various languages, while English commonly uses discriminator.

In electronics and communications, a diskriminator refers to a device used in FM demodulation to convert instantaneous frequency changes into a varying voltage that can drive audio stages. Classic implementations include slope discriminators and ratio detectors. The output is intended to be proportional to the instantaneous frequency deviation, enabling recovery of the original modulating signal. Design considerations for discriminators include linearity, drift, noise, and impedance matching, and they are typically found as part of the demodulator chain in FM receivers.

In machine learning, a discriminator is the component of a generative adversarial network (GAN) that learns to distinguish real data from data produced by a generator. The discriminator is trained to assign high probability to real samples and low probability to fake ones, guiding the generator through the adversarial objective. Common architectures include convolutional neural networks and multilayer perceptrons, with training framed as a minimax optimization problem.

In statistics and pattern recognition, discriminant analysis uses discriminant functions to classify observations into predefined groups. The term also appears in broader linguistic and philosophical contexts to describe criteria or mechanisms for making distinctions.

Overall, diskriminator or discriminator denotes both a theoretical concept of distinction and a practical device or model used to separate or identify signals, data, or categories across disciplines.