FIELD: data processing.
SUBSTANCE: invention relates to a method and a device for neuromorphic data processing. In the method, when power is supplied to a field programmable logic device (FPLD), comprising a neural network unit realizing neural network selection for the current task solution, which satisfies selection criteria, configuration of said neural network unit is downloaded from configuration memory, which based on analysis of input data of solved task and according to the above-specified criteria selects from the configuration memory corresponding neural network and loads it into the memory of the FPLD, providing dynamic reconfiguration of the FPLD, the loaded neural network performs data processing to solve the current task, and if the solution of the current task received by the loaded neural network does not satisfy predetermined criteria, then the neural network unit performs dynamic reconfiguration of the FPLD, loading from configuration memory image of another neural network, which satisfies current operating conditions, providing full or partial dynamic reconfiguration.
EFFECT: technical result is faster configuration of the FPLD.
2 cl, 6 dwg
