Model parameter learning device, control device, and model parameter learning method
AssigneesHonda Motor Co., Ltd., Kyoto University
InventorsKosuke Nakanishi, Yuji Yasui, Wataru Sasaki, Shin Ishii
Life of this patent
Estimating the term…
Abstract
Provided is a model parameter learning device and the like capable of learning model parameters such that the influence of a noise in input data can be suppressed. A model parameter learning device (1) alternately carries out first learning processing for learning model parameters W1, b1, W2 and b2 such that an error between data Xout and data Xorg is minimized, and second learning processing for learning model parameters W1, b1, Wm, bm, Wq and bq such that a loss function LAE is minimized.
Claims18
A model parameter learning device comprising:
- a characteristic data acquiring unit which acquires characteristic data, which is data of a parameter indicating at least one of a physical characteristic and a visual characteristic of an object;
- a first noise-added data creating unit which creates first noise-added data by adding predetermined first noise data to the characteristic data;
- a model data acquiring unit which has a first model and a second model, and inputs output data of the first model to the second model and acquires second model data output from the second model in a case where the first noise-added data is input to the first model;
- a teacher data acquiring unit which acquires teacher data corresponding to the characteristic data;
- and
- a first learning unit which carries out first learning processing for learning a model parameter of at least one of the first model and the second model such that an error between the second model data and the teacher data is minimized.
A model parameter learning method comprising steps of:
- acquiring characteristic data, which is the data of a parameter indicating at least one of a physical characteristic and a visual characteristic of an object;
- creating first noise-added data by adding predetermined first noise data to the characteristic data;
- inputting the first noise-added data to a first model;
- inputting output data of the first model to a second model;
- acquiring second model data, which is output data of the second model;
- acquiring teacher data corresponding to the characteristic data;
- and
- carrying out first learning processing for learning a model parameter of at least one of the first model and the second model such that an error between the second model data and the teacher data is minimized.
Dependencies are inferred from the claim text.