| 摘要 |
1,176,801. Computer programming. J. F. CANGUILHEM. 30 Dec., 1966 [5 Jan., 1966], No. 58447/66. Heading G4A. A computer is programmed to perform "dimensional synthesis" on a plurality of sets of values representing points or vectors in a multi-dimensional space, the synthesis proceeding according to one or more of six subprogrammes which calculate quantities (specified in the claim) concerned with respectively: (1) co-ordinate-scaling using standard deviation, a diagonal value also being calculated, (2) co-ordinate-scaling using range of variation, a diagonal value also being calculated, (3) value of a target quantity in absolute form, for a single entity, (4) as (3) but not in absolute form, (5) value of a target quantity in absolute form, for a group of entities, using vector addition in effect, (6) use of second-order moments for obtaining the value of a target quantity for a group of entities. Entities (e.g. firms) are each plotted as a point in a multidimensional space according to the values they have for certain factors (e.g. manpower, turnover, &c.), each factor corresponding to a respective dimension. The value of a target quantity (e.g. prosperity) for each entity is then given by the (magnitude of the) position vector of the point or the distance of the point from a point representing the optimum. This arises because each dimension is scaled in accordance with the relative importance of the factor towards the target quantity, by taking the magnitude of the unit vector for that dimension proportional to the difference of the largest and smallest values of the corresponding factor (over all entities) divided by the product of hierarchy coefficient representing the importance and a common scale factor. The difference mentioned may be replaced by the standard deviation of the values of the factor. The value of the target quantity may be expressed in absolute form viz. as a percentage of the largest diagonal of the parallelopiped whose sides are the ranges of variation of the respective factors (as found among the entities actually plotted, i.e. not possible variations). The sides of the parallelopiped may alternatively be the standard deviations of the respective factors. The value of the target quantity for the group of entities as a whole (or for a subset of them) may, depending on the nature of the target quantity, be obtained by vector addition of the position vectors of the points representing the entities, or by the principle of moments. In the second case, the target quantity for the group is that of an imaginary entity the importance of which is equal to the sum of the importances of the separate entities, and the second order moment of the importance of which (about the optimum point previously mentioned) is equal to the sum of such moments for the separate entities (the importance of an entity being considered to reside at the point representing the entity). Just as the target quantity can be calculated for a group of entities from its values for the entities as above, so its value for a set of groups can in the same way be calculated from its values for the groups, and so on. Entities can be ordered for desirability according to their distance in the space from the optimum point, or grouped for similarity according to their closeness to each other in the space. Optimization problems can be solved by minimizing the lengths of the trajectories traced out by points in the space as conditions vary. Sub-programmes (procedures) for mathematical calculations required above, and example programmes utilizing them for particular problems, are given in french ALGOL. |
