Computer simulations of learning in neural systems

Recent experiments have shown that, in some cases, strengths of synaptic ties are being modified in learning. However, it is not known what the rules that control those modifications are, especially what determines which synapses will be modified and which will remain unchanged during a learning episode. Two postulated rules that may solve that problem are introduced. To check their effectiveness, the rules are tested in many computer models that simulate learning in neural systems. The simulations demonstrate that, theoretically, the two postulated rules are effective in organizing the synaptic changes. If they are found to also exist in biological systems, these postulated rules may be an important element in the learning process.