Training Convolutional Neural Networks with Competitive Hebbian Learning Approaches

AbstractWe explore competitive Hebbian learning strategies to train feature detectors in Convolutional Neural Networks (CNNs), without supervision. We consider variants of the Winner-Takes-All (WTA) strategy explored previous works, i.e. k-WTA, e-soft-WTA and p-soft-WTA, performing experiments on different object recognition datasets. Results suggest that approaches are effective early extraction layers, or re-train higher layers a pre-trained network, with soft competition generally better than other this work. Our findings encourage path cooperation between neuroscience computer science towards deeper investigation biologically inspired principles.KeywordsNeural networksMachine learningHebbian learningCompetitive learningComputer visionBiologically