论文信息 - Self-Teaching Through Correlated Input

Self-Teaching Through Correlated Input

Previous work has shown that competitive learning coupled with a top-down teaching signal can produce compact invariant representations. In this paper we show that such a teaching signal can be derived internally from correlations between input patterns to two or more converging processing streams with feedback. Such correlations arise naturally from the structure present in natural environments. We demonstrate this process on two small but computationally difficult problems. We hypothesize that the correlations between and within sensory systems enable the learning of invariant properties.

Dana H. Ballard | Virginia R. de Sa

[1] V. de Sa,et al. Top-down teaching enables task-relevant classification with competitive learning , 1992, [Proceedings 1992] IJCNN International Joint Conference on Neural Networks.

[2] E. Rolls. The representation and storage of information in neural networks in the primate cerebral cortex and hippocampus , 1989 .

[3] E. Bienenstock,et al. Theory for the development of neuron selectivity: orientation specificity and binocular interaction in visual cortex , 1982, The Journal of neuroscience : the official journal of the Society for Neuroscience.

[4] Stephen Grossberg,et al. ARTMAP: supervised real-time learning and classification of nonstationary data by a self-organizing neural network , 1991, [1991 Proceedings] IEEE Conference on Neural Networks for Ocean Engineering.

[5] Paul Munro. Self-supervised Learning of Concepts by Single Units and "Weakly Local" Representations , 1988 .

[6] T. Kohonen,et al. Statistical pattern recognition with neural networks: benchmarking studies , 1988, IEEE 1988 International Conference on Neural Networks.