Inference apparatus, inference method, program and inference system

Abstract

To improve inference accuracy when using plural machine learning models.SOLUTION: An inference apparatus according to this technology includes: a first inference processing unit for executing inference by a first machine learning model; and a second inference processing unit for executing inference by a second machine learning model using the inference result from the first inference processing unit. The number of quantization bits of parameters used in the first machine learning model is greater than the quantization bit number of parameters used in the second machine learning model.SELECTED DRAWING: Figure 5

Images