Learning spiking neural network
Given a fully connected double-layer network two patterns are shown. This video is about supervised training.
The first pattern is white and the second is multi-colored: the synapses that connect the two patterns are strengthened and weakened until the first layer can learn to replicate the contents of the second. When sufficient precision is achieved, it is possible to block the synapses (color red) and eliminate the pattern to be learned (the multicolour); the latter is removed and the network is able to reproduce (depending on the input pattern and the value reached by the synapses) the content of the second. When the block is removed and nothing exists in the second pattern, the synapses become weak and the memory of the second pattern is erased
Видео Learning spiking neural network канала MultiNeurons
The first pattern is white and the second is multi-colored: the synapses that connect the two patterns are strengthened and weakened until the first layer can learn to replicate the contents of the second. When sufficient precision is achieved, it is possible to block the synapses (color red) and eliminate the pattern to be learned (the multicolour); the latter is removed and the network is able to reproduce (depending on the input pattern and the value reached by the synapses) the content of the second. When the block is removed and nothing exists in the second pattern, the synapses become weak and the memory of the second pattern is erased
Видео Learning spiking neural network канала MultiNeurons
Показать
Комментарии отсутствуют
Информация о видео
Другие видео канала
Adapting synaptic weights (2)Neural swimming poolA bunch of neuronsChemical / Mechanic dynamical systemColumn action potential waveSpiking neural network with conduction delayCluster of neuronsGroups of neuronsSpiking neuron phase portraitprtotype of a neuronSelf organizationtestAdaptive synaptic weightsneutoolfractalnetwork of neuronsAmorphic spiking neural networkSpiking neural networkLarge multilayer spiking neural networkSelf organization