The basins of attraction of a new Hopfield learning rule.

The nature of the basins of attraction of a Hopfield network is as important as the capacity. Here a new learning rule is re-introduced. This learning rule has a higher capacity than that of the Hebb rule, and still keeps important functionality, such as incrementality and locality, which the pseudo-inverse lacks. However the basins of attraction of the fixed points of this learning rule have not yet been studied. Three important characteristics of basins of attraction are considered: indirect and direct basins of attraction, distribution of sizes of basins of attraction and the shape of the basins of attraction. The results for the new learning rule are compared with those of the Hebb rule. The size of direct and indirect basins of attractions are generally larger for the new rule than for the Hebb rule, the distribution of sizes is more even, and the shape of the basins more round.