We use modified HNN for face recognition. To validate our algorithm, randomly generated distorted faces with 5%, 10% and so on, up to 45% distortion are presented to the network. For each distortion percentage the network is presented with 500 different distorted images. We observe the number of times the network converges to the correct fundamental face when presented with the distorted version of the fundamental face. Fig. 3 shows images of face 3 with 10% distortion face I with 25% distortion and face 2 with 35% distortion and respective retrieved faces. Table I summarized the response of the network for 7 faces (60x60 pixels). The results show at least 82.8% retrieval by CHNN and 63% retrieval by HNN for up to 45% distortion for 7 faces (60x60). For ULRI (10x10
pixels), our results show at least 72.8% retrieval by CHNN and at least 59.4% retrieval by HNN. As the distortion increases, the retrieval algorithm of CHNN performed better than HNN in terms of successful retrieval.