Abstract
Recent psychological and neurological evidence suggests that biological object recognition is a process of matching sensed images to stored iconic memories. This paper presents a partial implementation of (our interpretation of) Kosslyn’s biological vision model, with a control system added to it. We then show how reinforcement learning can be used to control and optimize recognition in an unsupervised learning mode, where the result of image matching is used as the reward signal to optimize earlier stages of processing.
Cite
CITATION STYLE
Draper, B. A., & Baek, K. (2000). Unsupervised learning of biologically plausible object recognition strategies. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 1811, pp. 238–247). Springer Verlag. https://doi.org/10.1007/3-540-45482-9_23
Register to see more suggestions
Mendeley helps you to discover research relevant for your work.
