位移不变的视觉模式在线学习和识别
Shift Invariant Visual Pattern Learning and Recognition
-
摘要: 提出了一种二维模式识别的神经网络模型ART-Cognizer,该模型采用新认知机的 结构和自适应共振网络的自顶向下的注意和匹配机制,因此对模式在视野中的缩放和平移有 容忍能力,并对自组织学习过程有自稳定能力,它可以在线学习新的模式而不损伤其原有记 忆.Abstract: This paper outlines a self-organizing neural network model, called ART-Cognizer, that automatically learns to classify visual patterns into recognition categories. It has the ability of tolerance to shift and scale variance of paterns, and the ability of self-stabilizing the self-organizing learning process, hence it can learn new patterns on-line without degrading its corpus of memories.
-
Key words:
- Visual pattern recognition /
- neocognitron /
- adaptive resonance theory
计量
- 文章访问数: 2901
- HTML全文浏览量: 71
- PDF下载量: 892
- 被引次数: 0