The plasma position and shape on the COMPASS-D tokamak have been controlled simultaneously with a 75-kHz bandwidth, hard-wired, real-time neural network. The primary network operates with up to 48 selected magnetic inputs and has been used in the vertical position control loop to control the position of the upper edge of the plasma at the radius of a reciprocating Langmuir probe and to keep this constant during a programmed shape sequence. One of the main advantages of neural networks is their ability to combine signals from different types of diagnostics. Two coupled networks are now in use on COMPASS-D. A dedicated soft-X-ray network has been created with inputs from 16 vertical and 16 horizontal camera channels. With just four hidden units, it is able to accurately determine three output signals defining the plasma core radius, vertical position, and elongation. These signals are fed to the primary network along with selected magnetic inputs and four poloidal field coil control current inputs. The core data are expected to help characterize the equilibrium by providing information on the Shafranov shift and gradient of elongation, related to the equilibrium parameters p and li. This network, with 15 hidden units, is able to define 10 outputs capable of giving a parameterized display of the plasma boundary. This paper describes results from several networks trained on various combinations of inputs with (a) simulated inputs and output values, where the precision of the network can be tested; (b) experimental inputs and calculated output values, where operational precision can be tested; and (c) hardware networks, where real-time performance can be tested. The results confirm that the neural network method is capable of giving excellent precision in tokamak boundary reconstruction but that the necessary accuracy in the experimental inputs for this task is not easily achieved.