{"title": "SEXNET: A Neural Network Identifies Sex From Human Faces", "book": "Advances in Neural Information Processing Systems", "page_first": 572, "page_last": 577, "abstract": null, "full_text": "SEXNET: A NEURAL NETWORK \n\nIDENTIFIES SEX FROM HUMAN FACES \n\nB.A. Golomb, D.T. Lawrence, and T.J. Sejnowski \n\nThe Salk Institute \n\n10010 N. Torrey Pines Rd. \n\nLa Jolla, CA 92037 \n\nAbstract \n\nSex identification in animals has biological importance. Humans are good \nat making this determination visually, but machines have not matched \nthis ability. A neural network was trained to discriminate sex in human \nfaces, and performed as well as humans on a set of 90 exemplars. Images \nsampled at 30x30 were compressed using a 900x40x900 fully-connected \nback-propagation network; activities of hidden units served as input to a \nback-propagation \"SexNet\" trained to produce values of 1 for male and \no for female faces. The network's average error rate of 8.1% compared \nfavorably to humans, who averaged 11.6%. Some SexNet errors mimicked \nthose of humans. \n\n1 \n\nINTRODUCTION \n\nPeople can capably tell if a human face is male or female. Recognizing the sex of \nconspecifics is important. ''''hile some animals use pheromones to recognize sex, in \nhumans this task is primarily visual. How is sex recognized from faces? By and \nlarge we are unable to say. Although certain features are nearly pathognomonic for \none sex or the other (facial hair for men, makeup or certain hairstyles for women), \neven in the absence of these cues the determination is made; and even in their \npresence, other cues may override. \n\nSex-recognition in faces is thus a. prototypical pattern recognition task of the sort \nat which humans excel, but which has vexed traditional AI. It appea.rs to follow \nno simple algorithm, and indeed is modifiable according to fashion (makeup, hair \netc). While ambiguous cases exist, for which we must appeal to other cues such as \nphysical build (if visible), voice patterns (if audible), and mannerisms, humans are \n\n572 \n\n\f", "award": [], "sourceid": 405, "authors": [{"given_name": "B.A.", "family_name": "Golomb", "institution": null}, {"given_name": "D.T.", "family_name": "Lawrence", "institution": null}, {"given_name": "T.J.", "family_name": "Sejnowski", "institution": null}]}