基于稀疏RAM的逼近型N-TUPLE神经网络模型

Approximate Type N-tuple Network with Sparse RAM

将单层查找感知器 (SL L UP)中的 RAM式神经元替换为稀疏 RAM神经元 ,获得了一个基于稀疏 RAM的逼近型 N- tuple神经网络新模型 .新模型不仅可以缓和原 SL L UP的存储容量易于饱和的问题 ,而且使网络的表示更灵活、更广泛 ,更重要的是使得 N- tuple类型的网络直接处理大维数的样本、大样本集类的数据真正成为可能 ,而不需要二值化样本 .

An approximate N tuple network with sparse RAM has been obtained by replacing RAM neurons in single layer lookup perceptrons(SLLUP) with sparse RAM ones. The new model not only alleviates memory saturation, but also behaves flexibly and extensively. It is more important that the improved N tuple version makes dealing directly with high dimension samples and with large dataset size possible and need no encoding to binary one. Experiments confirm the improved model reasonable and effective.