Small Codes for Big Signals

  • David Forsyth


This chapter explores a different kind of use of neural networks. Rather than classifying or detecting patterns directly, we try to build low dimensional representations of high dimensional signals. The simplest reason to do so is to build a map of a dataset. We’ve already seen one procedure for doing so. It turns out that procedure has problems; this chapter starts with two alternative procedures. These are useful in their own right for mapping datasets.

Copyright information

© Springer Nature Switzerland AG 2019

Authors and Affiliations

  • David Forsyth
    • 1
  1. 1.Computer Science DepartmentUniversity of Illinois Urbana ChampaignUrbanaUSA

Personalised recommendations