Hardware Realisable Learning Algorithms
暂无分享,去创建一个
Learning algorithms are developed which are implementable by means of RAM hardware. In particular probabilistic RAMs are used to implement unsupervised reward and associative learning algorithms. These are considered after an introductory survey of supervised gradient descent learning in pRAM form. Such learning is quite effective for certain hard problems, but reward and associative learning are shown to be more advantageous due to their local nature. The development of topological maps in nets of pRAMs is also shown to occur.
[1] D. Gorse,et al. AN ANALYSIS OF NOISY RAM AND NEURAL NETS , 1989 .
[2] I. Aleksander. The logic of connectionist systems , 1989 .