Sie befinden Sich nicht im Netzwerk der Universität Paderborn. Der Zugriff auf elektronische Ressourcen ist gegebenenfalls nur via VPN oder Shibboleth (DFN-AAI) möglich. mehr Informationen...
Performance Evaluation Dimensional Reduction Techniques for Image Classification
Ist Teil von
2021 International Conference on Machine Learning and Cybernetics (ICMLC), 2021, p.1-6
Ort / Verlag
IEEE
Erscheinungsjahr
2021
Quelle
IEEE Xplore
Beschreibungen/Notizen
There is a general consensus that obtaining a highly optimized and accurate representation of incoming signals is of paramount importance for both artificial intelligence systems and biological brains. For artificial systems, unlike biological systems, extracting meaningful models from high dimensional data to operate for real world tasks is still largely open issue. In this study, we introduce simple frameworks for dimensional reduction task : Hierarchical Temporal Memory (HTM), Principal Components Analysis (PCA) and fusing two frameworks. The first one has a neuroscience foundation and the second one is from unsupervised machine learning. By conducting image classification task, we analyze the performance of each framework and potential for synergistic interactions between two frameworks for possible performance improvements. Our experimental outcomes demonstrate that standalone PCA performs the best in the most cases. However, properly integrated usage of fused framework achieves similar or better image classification accuracy with significantly decreased classification time than the standalone framework.