Resumen de Weightless Neuron Models
Edson Costa Barros Carvalho Filho de
The study of connectionist processing and artificial neural networks has generated a wide diversity of models, both in terms of the nature of the computational nodes employed and the structural architecture for implementation. While many models are based on nodes which compute the familiar sum-of-products function, and alternative class of neural networks use adaptive Boolean logic elements for the processing node. In principle, such networks offer a number of advantages relating particularly to their potential for high-speed processing, simplicity of implementation in VLSI, tractability of analytical studies and real world practical tasks. This class of nodes is known as weightless or RAM-Based neurons because they present a structure similar to a look-up table or Random-Access-Memory. This paper review the Random-Access-Memory (RAM) Model [1], the Probabilistic Logic Neuron Model (PLN) [5], the Multi-Valued PLN[7] and the Goal-Seeking Model (GSN)[4].
