| 摘要 |
PURPOSE: To generate optimum learning data effective for learning and to efficiently and accurately perform the learning. CONSTITUTION: A conditional probability estimation part 2 repeats the updating of a parameterθuntil an optimum parameter is obtained and obtains a final parameter <θ> by calculating the parameter <θ> from learning data selected by a learning data selection part 5, supplying the calculated parameter to the learning data selection part 5, letting the learning data selection part 5 select the new learning data based on the parameter <θ> and calculating the new parameter <θ> from the selected new learning data.
|
