There is Incremental Learning as one of the chaotic neural network's learnings. In Incremental Learning, the network learns by feeding several patterns to it. The results of learnings probably change by differences of the input patterns. This study especially examined how the learning is influenced by the binary ratio of 1 and -1 in input patterns.
At first, the ratio of 1 was increased from 10% to 90% by 5% and random patterns were generated in each ratio. And these patterns were input to the network which consist of 100 neurons and were learned. As a result of first experiment, the ratio of 10% or 90% gave larger maximum number of complete learning than the ratio of 50%. Therefore, it's considered that high correlated patterns are more easy to be learned than not high correlated ones.
Next, patterns were also generated and learned in the network which consist of 200 or 300 neurons. As a result of second experiment, the number of learned patterns was larger than the one in experiment of 100 neurons in any ratio. Big increases were especially seen in ratios of 10% and 90%. For maximum number of complete learning, differences in ratios were hardly seen.
As the result of two experiments, it's considered that the binaly ratio of the input patterns certainly influences the learning. But there is the value which wasn't influenced by the binary ratio in some network condition.