Computing apparatus, computing method, storage medium, and table generating apparatus

A computing method for activation function computation of a neural network, comprising:

• retaining, by a computing apparatus, an approximation table that approximately represents an activation function of a neural network, the approximation table mapping between a plurality of discrete input samples of the activation function and a plurality of output samples, the plurality of output samples respectively corresponding to the plurality of input samples;
• and
• converting an input value of the activation function computation to an output value using the approximation table when the activation function is selected for the activation function computation,
• wherein the plurality of input samples of the approximation table are set such that input samples more distant from a reference point in the domain of the activation function have a larger neighboring sample interval, and
• wherein the neighboring sample interval is set such that the neighboring sample interval increases, from a basic sample interval at the reference point as a baseline, by a constant ratio or a constant increment with increasing distance from the reference point.

A non-transitory computer-readable storage medium having stored therein a computer program which causes a processor of a computing apparatus that performs activation function computation of a neural network to perform:

• retaining an approximation table that approximately represents an activation function of the neural network, the approximation table mapping between a plurality of discrete input samples of the activation function and a plurality of output samples, the plurality of output samples respectively corresponding to the plurality of input samples;
• and
• converting an input value of the activation function computation to an output value using the approximation table when the activation function is selected for the activation function computation,
• wherein the plurality of input samples of the approximation table are set such that input samples more distant from a reference point in the domain of the activation function have a larger neighboring sample interval, and
• wherein the neighboring sample interval is set such that the neighboring sample interval increases, from a basic sample interval at the reference point as a baseline, by a constant ratio or a constant increment with increasing distance from the reference point.

A table generating apparatus comprising:

• a receiving unit configured to receive a designation of an activation function of a neural network;
• and
• a generation unit configured to generate an approximation table that approximately represents the designated activation function, the approximation table mapping between a plurality of discrete input samples of the activation function and a plurality of output samples, the plurality of output samples respectively corresponding to the plurality of input samples,
• wherein the generation unit is configured to set the plurality of input samples of the approximation table such that input samples more distant from a reference point in the domain of the activation function have a larger neighboring sample interval, and
• wherein the neighboring sample interval is set such that the neighboring sample interval increases, from a basic sample interval at the reference point as a baseline, by a constant ratio or a constant increment with increasing distance from the reference point.