The amount of information required to memorize paths

This paper proposes a method to measure the amount of information required to memorize a path on a road network. First, an experimental model of a walker is proposed, which takes a roll of a path recognizer instead of a real person. The path recognition cost is defined as the amount of storage needed to encode a test which is offered to the recognizer. The amount is measured as the number of bits, actually. Then the minimal storage cost path between two points is defined and a quasi minimal storage cost tree is discussed. The algorithms to compute these are shown. Then the distances and storage cost are computed and compared on both this tree and the shortest distance tree on an actual road network. Finally, the expectation of the storage cost is examined on some random network models. [In Japanese.]