Accounting for class hierarchy in object classification using Siamese neural networksстатьяИсследовательская статья
Информация о цитировании статьи получена из
Scopus
Статья опубликована в журнале из списка Web of Science и/или Scopus
Дата последнего поиска статьи во внешних источниках: 1 мая 2024 г.
Аннотация:Siamese neural networks are an effective architecture for automatic construction of vector representations of objects, by whose comparison it is possible to solve the classification problem. The mentioned approach utilizes the training selection more effectively; it is capable of distinguishing classes by a small number of samples, and it can learn with a dynamic number of classes. We propose an improved method of tuning the Siamese neural networks for solving the classification problem, that uses information about the class hierarchy. The advantage of the mentioned method is demonstrated in examples of image and text classifications.