Hybrid clustering-classification neural network in the medical diagnostics of reactive arthritis

• DOI: 10.48550/arxiv.1610.07857
• Publication Date: 2016-01-01
• Authors: Yevgeniy Bodyanskiy; Olena Vynokurova; Volodymyr Savvo; Tatiana Tverdokhlib; Павло Мулеса
• Keywords: Cosine similarity

The hybrid clustering-classification neural network is proposed. This allows increasing a quality of information processing under the condition overlapping classes due to rational choice learning rate parameter and introducing special procedure fuzzy reasoning in clustering process, which occurs both with an external signal (supervised) without one (unsupervised). As similarity measure neighborhood function or membership one, cosine structures are used, allow provide high flexibility self-learning-learning process some new useful properties. Many realized experiments have confirmed efficiency proposed network; also, this was used for solving diagnostics task reactive arthritis.