A rotation-invariant neocognitron

Satoh, Shunji; Kuroiwa, Jousuke; Aso, Hirotomo; Miyake, Shōgo

doi:10.1002/(sici)1520-684x(199904)30:4<31::aid-scj4>3.0.co;2-7

Cited by 6 publications

(1 citation statement)

References 4 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The x, y components of the summed vector (population vector) are calculated by Eq. (17). Then, the orientation of the population vector I is calculated by Eq.…”

Section: Recollection Processmentioning

confidence: 99%

Evaluation of recognition ability and inside parameters for a rotation spreading associative neural network

Nakamura¹,

Yoshikawa²

2001

Electron Comm Jpn Pt III

View full text Add to dashboard Cite

SUMMARYFocusing on the parietal cortex, we have previously proposed a spreading associative neural network that could simultaneously recognize the orientation and shape of a binary character and a gray-scaled human face. We evaluated the recognition performance of the spreading associative neural network when the number of learning directions, number of spreading directions, and spreading range were varied. We searched for the optimal parameters of this neural net to obtain the best recognition performance based on the present evaluation. In general, the recognition performance was improved when the number of spreading directions was greater than (or equal to) the number of learning directions and the spreading range equaled the nonlearned range. Giving consideration to the results of recognition characteristics and learning and recognition time, it is suggested that the number of learning directions should be equal to the number of spreading directions.

show abstract

“…The x, y components of the summed vector (population vector) are calculated by Eq. (17). Then, the orientation of the population vector I is calculated by Eq.…”

Section: Recollection Processmentioning

confidence: 99%