Abstract: A recognition device which judges whether a target is identical with a predetermined reference, comprising: a holding unit which holds the target; multiple deformation units which deform the target held by the holding unit with at least one degree of flexibility in deformation; multiple deformed amount estimation units which correspond to the deformation units in a one-to-one relationship and estimate a deformed amount of the target from the reference with respect to the flexibility in deformation according to the corresponding deformation unit; an estimated error evaluation unit which evaluates an estimated error of the deformed amount estimated by the deformed amount estimation unit; an adjustment unit which operates any of the deformation units with precedence according to the estimated error evaluated by the estimated error evaluation unit; a similarity calculation unit which calculates a similarity between the reference and the target which is deformed by the deformation unit operated with precedence by

Abstract: A data classifier classifies a plurality of input pattern data into one or more clusters. For each pattern data, a cluster to which the pattern data belongs is provisionally determined. For each cluster, a predetermined correlation value is calculated between one or more pattern data belonging to the cluster and observational pattern data which is a target to be classified into a cluster. A cluster to which the observational pattern data belongs is determined based on the correlation values.

Abstract: A data classifier performs a data classification process using prototypes classified into clusters. A prototype map is formed using mapping means and clustering means. The mapping means forms, through learning, a prototype map by adjusting coupling weights between a plurality of prototypes provided in a map space based on a plurality of input data. The clustering means calculates a predetermined measure between the prototypes and classifies the prototypes into a plurality of clusters based on the measure.

Abstract: A pattern recognition method and apparatus decrease the amount of computation for pattern recognition and adapts flexibly to an increase and a change in learning samples. Learning is made beforehand on base vectors in a subspace of each category and a kernel function. Pattern data to be recognized is inputted, and projection of an input pattern to a nonlinear subspace of each category is decided. Based on the decided projection, a Euclidean distance or an evaluation value related to each category is calculated from the property of the kernel function, and is compared with a threshold value. If a category for which the evaluation value is below the threshold value exists, a category for which the evaluation value is the smallest is outputted as recognition results. If there is no category for which the evaluation value is below the threshold value, a teaching signal is inputted for additional learning.

Abstract: An image sensor which has a simple construction, and can read a voltage value, not a current value, based on the quantity of light received by the photo diode, and can read charge of the light sensing elements every block. A plurality of light sensing elements each consisting of a photo diode and a blocking diode, make up a block. A plurality of blocks are arrayed. The blocking diodes are connected to a shift register for transferring a bias voltage thereto. The lead wires from the photo diodes are connected to common signal lines. The common signal lines are respectively connected to load capacitors. Potential caused in each common signal line by the charge stored in the corresponding load capacitor is read out.