Technologies for optimized machines learn training

摘要

Technologies for optimizing of machines learn training include a computing device for training a machines learn network with a training algorithm, which is configured with configuration parameters. The computing device, a large number of training can be carried out in parallel into the dancing. The computing device, from the training a time series of part precise values to. Each part precise value indicates a machines learn network accuracy in the case of an associated training iteration to. The computing device, the configuration parameters in a feedforward neural network an for producing a representation and releases the representation in a feedback neural network. The computing device, the feed forward modulation trained neural network and the fed-back neural network against the part precise values. The computing device can be optimized, the feed forward modulation neural network and the fed-back neural network for determining an optimized configuration parameters. The optimized configuration parameters, the training time for reaching a predetermined accuracy level minimize. Other embodiments are described and claimed.