摘要 |
<p>PURPOSE: To provide the data classifying device which functions fast with high precision according to the whole data and the distribution of individual data when >=2 kinds of data having different properties are classified by the properties. CONSTITUTION: This data classifying device consists of a learning data storage part 10, a reference vector selection part 11, a reference vector storage part 12, a closest reference vector search part 13, a corrected reference vector selection part 14, a reference vector correction part 15, a learning data classification part 16, a category determination part 17, a duplication and deletion decision part 20, and a duplication and deletion part 21 for matching an initial value set in consideration of the number of reference vectors and their categories with learning data, a learning vector quantization part 18 which learns with the algorithm of learning vector quantization, a convergence decision part 22 which ends learning at the least learning frequency for obtaining necessary precision and speeds up the learning, and an estimation part 19.</p> |