The Innovative Systems Design Group at SGS-Thomson Microelectronics has been exploring analog implementations of neural network architectures with special emphasis on efficiency, precision, and large-scale implementation. We are developing three large-scale analog VLSI chips, all of which work on image-processing problems. Because of their efficiency and regularity, analog computing arrays form the basis of our designs, which use current, charge, and conductance computing modes. We have also investigated the use of floating-gate flash-EEPROM devices for both nonvolatile analog storage and computation.