Reinforcement learning for homeostatic endogenous variables

An announcing system for providing a family of recorded announcements through telephone central office equipment to one or more telephone subscriber lines. Each announcement selectively includes weather, temperature, or other periodic and aperiodic information segments. Each announcement is changeable in length from a long cycle which includes a full advertising segment to a short cycle in which a major portion or all of the advertising segment is deleted so that the time required for an announcement is shortened. The length of the announcement cycle is shortened in response to the number of subscriber lines waiting to be connected to the announcing system being in excess of the number of subscriber lines which can be connected to the announcing system by the telephone central office equipment. However, the announcing system also provides for deleting segments of an announcement which are in error without changing the length of the announcement cycle or it being apparent that the announcement segment has been deleted. The weather segment of the announcement is determined by different three digit weather codes and by changing the positions of a plurality of weather announcing heads along a weather announcing drum in response to a plurality of resistances corresponding to the three digit weather codes.