𝔖 Bobbio Scriptorium
✦   LIBER   ✦

Neural ensembles: Simultaneous recognition of multiple 2-D visual objects

✍ Scribed by Malcolm R.J. Mcquoid


Publisher
Elsevier Science
Year
1993
Tongue
English
Weight
982 KB
Volume
6
Category
Article
ISSN
0893-6080

No coin nor oath required. For personal study only.

✦ Synopsis


Presented is a hierarchical neural network that using some highly simplified ideasfrom the neurophysiology ofthe visual cortex can simultaneously recognise multiple object s within the same field ofview without recourse to f eedback or a competitive learning type algorithm . It employs ensem bles of neurons that within each layer are isolated from one another and can adapt itselfeither with or without sup ervision, requiring only one forward pass for learning or recognition. The adaption algorithm for each ensemble or neuron is purely local and is independent from the system's output response. Experiments verified the network's ability for sim ultaneous multiple-object recognition and also gave an indication of its performan ce with noisy images. A method ofautomatic neuron threshold dropoff was found to give increased performance in the recognition ofnoisy image s.