| 摘要 |
A learning apparatus performs a learning process for a feed-forward multilayer neural network with supervised learning. The network includes an input layer, an output layer, and at least one hidden layer having at least one probing neuron that does not transfer an output to an uppermost layer side of the network. The learning apparatus includes a learning unit and a layer quantity adjusting unit. The learning unit performs a learning process by calculation of a cost derived by a cost function defined in the multilayer neural network using a training data set for supervised learning. The layer quantity adjusting unit removes at least one uppermost layer from the network based on the cost derived by the output from the probing neuron, and sets, as the output layer, the probing neuron in the uppermost layer of the remaining layers. |
| 主权项 |
1. A learning apparatus that performs a learning process for a feed-forward multilayer neural network with supervised learning,
the multilayer neural network comprising:
a plurality of layers configured by an input layer that is a lowermost layer of a layered structure of the multilayer neural network, an output layer that is an uppermost layer of the layered structure, and at least one hidden layer that is located between the input layer and the output layer,each of the layers including a given number of units that receive an input from the lowermost layer side, perform a predetermined calculation based on the input and a weight to produce an output, and transfer the output toward the uppermost layer side,the at least one hidden layer including at least one probing neuron that receives an input from the lowermost layer side, performs a predetermined calculation based on the input and a weight to produce an output, but does not transfer the output to the uppermost layer side, the learning apparatus comprising:
a learning unit that performs a learning process by calculation of a cost derived by a cost function defined in the multilayer neural network using a training data set for supervised learning; anda layer quantity adjusting unit that removes at least one uppermost layer from the multilayer neural network based on the cost derived by the output from the at least one probing neuron, and sets, as the output layer, the at least one probing neuron in the uppermost layer of the remaining layers. |
