Learning device and learning method, recognition device and recognition method, program, and storage medium

A recognition device comprising:

• a processor;
• and
• a memory storing program code executable by the processor to:
• apply a neural network to a plurality of element values of recognition target data to perform recognition on the recognition target data, recognition target data being image data or voice data,
• wherein the neural network has a plurality of learning cells each including a plurality of input nodes that perform predetermined weighting on each of the plurality of element values and an output node that sums the plurality of weighted element values and outputs the sum,
• wherein each of the plurality of learning cells is linked with any of a plurality of categories indicting training information,
• wherein the plurality of input nodes of each of the learning cells are configured such that each of the plurality of element values is input at a predetermined weight in accordance with a corresponding category, and
• wherein based on an output value of each of the learning cells and a category linked with each of the learning cells, the processor recognizes a category to which the recognition target data belongs.

A learning method of a learning device, comprising:

• training a neural network adapted to perform recognition based on image data or voice data, the neural network having a plurality of learning cells each including a plurality of input nodes to which a plurality of element values based on learning target data are input and an output node that sums the plurality of element values at respective predetermined weights for the plurality of element values and outputs the sum, including:
• inputting a plurality of element values based on the learning target data to the plurality of input nodes of the learning cells;
• and
• in accordance with an output value of each of the learning cells, updating weighting coefficients of the plurality of input nodes of the learning cells or adding a new learning cell to the neural network unit;
• and
• after training the neural network, applying the neural network to the image data or the voice data to perform the recognition on the image data or the voice data.

A recognition method using a recognition device, comprising:

• applying a neural network to perform recognition on recognition target data, the recognition target data being image data or voice data, the neural network having a plurality of learning cells, wherein each of the learning cells includes a plurality of input nodes to which a plurality of element values based on the recognition target data are input and an output node that sums the plurality of element values at predetermined weights for the plurality of element values and outputs the sum, and each of the learning cells is linked with any of a plurality of categories indicating training information, including:
• inputting a plurality of element values based on the recognition target data to the plurality of input nodes of the plurality of learning cells;
• and
• based on an output value of each of the learning cells and a category linked with each of the learning cells, recognizing a category to which the recognition target data belongs.

A non-transitory computer readable storage medium storing a program executed by a learning device to:

• train a neural network adapted to perform recognition based on image data or voice data, including:
• inputting a plurality of element values based on learning target data to the neural network,
• wherein the neural network has a plurality of learning cells each including a plurality of input nodes to which a plurality of element values based on the learning target data are input and an output node that sums the plurality of weighted element values and outputs the sum;
• and
• in accordance with an output value of each of the learning cells, updating weighting coefficients of the plurality of input nodes of each of the learning cells or adds a new learning cell to the neural network unit;
• and
• after the neural network has been trained, apply the neural network to the image data or the voice data to perform the recognition on the image data or the voice data.

A non-transitory computer readable storage medium storing a program executed by a recognition device:

• apply a neural network to perform recognition based on recognition target data, the recognition target data, the recognition target data being image data or voice data, the neural network having a plurality of learning cells, wherein each of the learning cells includes a plurality of input nodes to which a plurality of element values based on the recognition target data are input and an output node that sums the plurality of weighted element values and outputs the sum, and each of the learning cells is linked with any of a plurality of categories indicting training information, including:
• inputting a plurality of element values based on the recognition target data to the plurality of input nodes of the plurality of learning cells;
• and
• based on an output value of each of the learning cells and a category linked with each of the learning cells, recognizing a category to which the recognition target data belongs.