Error correcting network

Summary form only given. The authors propose a network that uses the simplest learning rules but exhibits a high storage capacity. This increase in storage capacity is achieved by introducing error correction on the output of traditional low-capacity networks. Experimental error-corrected recall indicated that the error-correcting network had the ability to recall very nearly all of the available information produced by a simple Hopfield-type network long after the Hopfield-type network had failed to recall any association by itself.<>