Output details
15 - General Engineering
University of Edinburgh (joint submission with Heriot-Watt University)
Minimizing the Effect of Process Mismatch in a Neuromorphic System Using Spike-Timing-Dependent Adaptation
This is the first demonstration of a biologically-derived, spike-correlation learning circuit that improves the performance of a pulse-coded CMOS chip (neural or, for example, a sigma-delta converter). Device mismatches cause chip performance to degrade. This method, using pre-existing correlations in the spiking signals to ameliorate transistor mismatches, will ultimately allow new forms of self-calibrating spiking circuits to be designed for nanoscale silicon. The work was funded by EP/C516583/1, £421k,2005-09 and has influenced the work of groups in Harvard-MIT (DoI: 10.1073/pnas.1106161108) and ETH/Zurich (DoI:10.1109/IJCNN.2012.6252636).
