- Full text PDF available (1)
Journals and Conferences
probabilistic computation applications show that this architecture is resilient to error rates on the order of trillion-plus failures-in-time (FITs, defined as the number of failures in 10 9 hours), while still delivering robust results with competitive performance.
An adaptive neural network architecture is proposed to perform segmentation in real time for mixed gray-level and binary pictures. In this approach the composite picture is divided into binary and image blocks based on a dichotomy measure computed by an adaptive neural net. A VLSI implementation of such a net in 2- mu m CMOS technology is described and… (More)