{"pk":31679,"title":"Tau Net: The Way to do is to be","subtitle":null,"abstract":"We describe a technique for automatically adapt-\ning to the rate of an incoming signal. We first\nbuild a model of the signal using a recurrent net-\nwork trained to predict the input at some delay,\nfor a \"typical\" rate of the signal. Then, fixing the\nweights of this network, we adapt the time con-\nstant T of the network using gradient descent,\nadapting the delay appropriately as well. W e\nhave found that on simple signals, the network\nadapts rapidly to new inputs varying in rate from\ntwice as fast as the original signal, d o w n to ten\ntimes as slow. So far our results are based on\nlinear rate changes. We discuss the possibilities\nof application to speech.","language":"eng","license":{"name":"","short_name":"","text":null,"url":""},"keywords":[],"section":"Submitted Presentations","is_remote":true,"remote_url":"https://escholarship.org/uc/item/01c4x1gs","frozenauthors":[{"first_name":"Garrison","middle_name":"W.","last_name":"Cottrel","name_suffix":"","institution":"University of California, San Diego","department":""},{"first_name":"Mai","middle_name":"","last_name":"Nguyen","name_suffix":"","institution":"University of California, San Diego","department":""},{"first_name":"Fu-Sheng","middle_name":"","last_name":"Tsung","name_suffix":"","institution":"University of California, San Diego","department":""}],"date_submitted":null,"date_accepted":null,"date_published":"1993-01-02T02:00:00+08:00","render_galley":null,"galleys":[{"label":"PDF","type":"pdf","path":"https://journalpub.escholarship.org/cognitivesciencesociety/article/31679/galley/22747/download/"}]}