An initial performance evaluation of unsupervised learning with ALIAS

Based on collective learning systems theory and a versatile general-purpose architecture for massively parallel networks of processors, a transputer-based parallel-processing image-processing engine comprising a three-layer hierarchical network of 32 learning cells and 33 nonlearning cells has been applied to a difficult image-processing task: the detection of anomalous features in otherwise normal images. Known as ALIAS (adaptive learning image analysis system), this engine is currently being constructed and tested. ALIAS is limited to the translation and scale-invariant detection of anomalies. Future enhancements will include rotational invariance as well as the automatic classification of images. An experiment with unsupervised learning indicates excellent detection of anomalies which are square sections of the image shifted horizontally or vertically with respect to the original image. Supervised learning, to be implemented in the near future, will allow ALIAS to be conditioned to accept or reject specific anomalous features (either normal or abnormal), as appropriate