Abstract
For many years the similarity of taste stimuli has been measured mainly by the determination of a correlation coefficient across the neurons activated by the stimuli. In a previous study, we developed a new method for analyzing gustatory neural activities, using artificial neural networks [1]. In the present study, we examined further the operation of these networks in regard to neural responses induced by a variety of taste stimuli, and the effect of “pruning” on the network output. Three-layer neural networks were trained by the back-propagation learning algorithm to classify the neural response patterns to the four basic taste qualities (1.0 M sucrose, 0.03 M HCl, 0.01 M quinine HCl, and 0.1 M NaCl). The networks had four output units representing the four basic taste qualities: sweet (S), sour (H), bitter (Q), and salt (N). The input units represented rat cortical neurons from which the response patterns (impulses/3 s) were recorded in a separate physiological experiment [2]. After training, the response patterns to test stimuli (0.02 M sodium saccharin, 0.3 M KCl, 0.3 M MgCl2, 0.1 M NaNO3, 0.01 M tartaric acid, 0.3 M CaCl2, 0.1 M monosodium L-glutamate [MSG] and 0.1 M inosine 5′-monophosphate [IMP]) were presented to the input units. For NaNO3, the networks produced large outputs (around 0.9) almost exclusively in the N unit, showing pure salt taste in the stimulus. Large and exclusive outputs were also produced in the H unit for KCl (around 0.7) and in the Q unit for tartaric acid (around 0.5). On the other hand, outputs suggesting mixed bitter and salt tastes were produced for CaCl2 and MgCl2. As to the similarity of the test stimuli to the four basic taste qualities, the neural networks presented a clearer and more definite result than the conventional correlation analysis. For MSG and IMP, the networks produced exclusive but small outputs (around 0.2) in the N unit, although some investigators classify these stimuli as an umami taste that is independent of the four basic taste qualities.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Disclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.