Reconfigurable Acousto-optic Processor for Artificial Neural Network Training
Outline
Project aims
Background Spatial light modulators
Background Acousto-optic effect
Background Acousto-optic devices
Background Calomel
Background The perceptron
Algorithm Matrix-vector multiplication
Experimental realisation
Experimental realisation (cont.)
PPT Slide
Experimental realisation Selective (spatial) frequency filtering
Experimental concerns Time-independent
Experimental concerns Time-dependent
Experimental concerns (cont.)
Perceptron learning
Perceptron learning Invariance of h due to weight-space normalisation
Recommendations
Conclusions
Future directions
Creating the system Representing negative weights
Creating the system Representing negative weights (cont.)
Email: tom.naughton@may.ie
Home Page: http://www.cs.may.ie/~tnaughton