| 摘要 |
<p>The invention relates to a device and to a method for learning sequential messages, said device and method using a neural network, said network comprising binary neurons, or beacons, distributed in C-clusters (23;), said clusters (23;) being arranged such that it is possible to identify, for a given cluster, a series of previous clusters according to a predefined order, and means for learning said neural network, comprising means for splitting a message to be learned into sub-messages to be learned, a means for activating, for one of said sub-messages to be learned, a single beacon in the on state in a current cluster, for one of said sub-messages to be learned, all of the other beacons of said current cluster being in the off state, means for creating connections between beacons, which activate, for said sub-message to be learned, connections between the beacon in the on state of said current cluster and the beacon in the on state of each of the clusters in the series of clusters preceding said current cluster, said connections being binary and unidirectional connections. A device and a method for decoding are also presented.</p> |
