Neural network circuit providing for operations with bit shifting and rounded values of weight information
AssigneeHitachi, Ltd.
InventorsToshiaki Nakamura, Teppei Hirotsu, Tatsuya Horiguchi
Life of this patent
Estimating the term…
Abstract
The present invention addresses the problem of implementing a neural network using a small-scale circuit by simplifying the multiplication of the input data by weight data. The neural network circuit according to the present invention is configured from: a means for multiplying input data by a rounded value of the mantissa part of weight data; a means for shifting the multiplication result by the number of bits of the rounded value; a means for adding the shifted result to the original input data; and a means for shifting the addition result by the number of bits of the exponent part of the weight.
Claims6
A neural network circuit comprising
- computer circuitry for providing a plurality of weight factors corresponding to a plurality of pieces of input data as information indicating a weight rounded value and an amount of bit shift and performing a weighting operation on the input data by multiplication with the weight rounded value and the bit shift, in neurons constituting a neural network,
- wherein the computer circuitry is configured to learn to obtain a weight factor of the plurality of weight factors using a floating-point operation, wherein the obtained weight factor is approximated to a weight factor obtained by the calculation Wa=S*(1+R*2−m)*2k, where S is a positive or negative sign of a weight, Wa is the weight factor, R is a rounded value of a weight and is an integer in a range of less than 0 and R<2m, and k is a bit shift number corresponding to the exponent of weight and is an integer.
A neural network circuit comprising:
- computer circuitry configured to
- multiply input data by a rounded value of a mantissa part of weight data;
- shift the multiplication result by the number of bits of the rounded value;
- add the shifted result to the original input data;
- and
- shift the addition result by the number of bits of an exponent part of the weight,
- wherein the computer circuitry is configured to learn to obtain a weight factor for the weight data using a floating-point operation, wherein the obtained weight factor is approximated to a weight factor obtained by the calculation Wa=S*(1+R*2−m)*2k, where S is a positive or negative sign of a weight, Wa is the weight factor, R is a rounded value of a weight and is an integer in a range of less than 0 and R<2m, and k is a bit shift number corresponding to the exponent of weight and is an integer.
Dependencies are inferred from the claim text.