A numerical technique is developed for the spatial localization of 1/f noise sources in AlSb/InAs high-electron-mobility transistors. The technique is based on the microscopic modeling of 1/f noise using the method of Langevin random sources, in which the noise sources are introduced nonuniformly throughout the device. An efficient algorithm is proposed for the determination of the spatial distribution of these noise sources and local (mesh point) values of the Hooge parameter. The algorithm takes advantage of the linearity between the terminal noise current and the power of the noise sources and uses an optimization approach based on linear least-squares minimizations to find the distribution from the experimental noise characteristics. The density of noise sources in the InAs channel is found to be much larger near the top interface than at the bottom interface. The local values of the Hooge parameter vary from 10-2 at the top to 3×10-4 at the bottom of the channel.