Patent · US Active

Residual quantization for neural networks

US11586883B2 · kind B2 · utility

1Cited by
2References
20Claims
0Family size

Assignee

Inventors

Key dates

Filing dateDec 14, 2018
Grant dateFeb 21, 2023
Priority date
Expiry dateDec 15, 2041

Classification

  • Technology area (CPC G)Physics
  • CPC primaryG06N3/045
  • WIPO fieldComputer technology
  • WIPO sectorElectrical engineering

Abstract

Methods and apparatus are disclosed for providing emulation of quantized precision operations in a neural network. In some examples, the quantized precision operations are performed in a block floating-point format where values of a tensor share a common exponent. Techniques for selecting higher precision or lower precision can be used based on a variety of input metrics. When converting to a quantized tensor, a residual tensor is produced. In one embodiment, an error value associated with converting from a normal-precision floating point number to the quantized tensor is used to determine whether to use the residual tensor in a dot product calculation. Using the residual tensor increases the precision of an output from a node. Selection of whether to use the residual tensor can depend on various input metrics including the error value, the layer number, the exponent value, the layer type, etc.

Source: USPTO / EPO open patent data. Objective bibliographic and citation counts.