Title of Research Output
Learning how to build a neural network model of the tactile periphery
Faculty
Science
Supervisor Name
Dr. Mark Daley
Keywords
Neural networks, receptive fields, model, mechanoreceptors
Description
First order neurons in the hairless skin of human hands have spatially complex receptive fields that allow for the detection of spatial details. These spatially complex receptive fields arise from the branching of mechanoreceptors, which converge and connect to first order neurons. This arrangement allows us to process our sensory environment through detecting the edge orientation of a touched object for instance, and do things like read braille.
These spatially complex receptive fields can studied by using a feedforward neural network to model the tactile periphery. By understanding the processing at the level of the tactile periphery, we can better understand the information that is fed to the CNS for further processing and coordination.
Compressed sampling allows a system to reconstruct a sparse signal with much fewer measurements when the sampling of a sparse input space is sufficiently random. It is our goal to incorporate compressed sampling into a neural network model of the tactile periphery to gain insight into the possible evolutionary processes that resulted in our ability to sense tactile details.
Acknowledgements
Special thanks to my supervisor Dr. Mark Daley for supporting this project and to the Computational Convergence Lab for taking me in, I am extremely grateful for everyone's mentorship. Thank you to Dr. Andrew Pruszynski for his crucial guidance and Dr. Charlie Zhao for beginning this project.
Document Type
Poster
Included in
Learning how to build a neural network model of the tactile periphery
First order neurons in the hairless skin of human hands have spatially complex receptive fields that allow for the detection of spatial details. These spatially complex receptive fields arise from the branching of mechanoreceptors, which converge and connect to first order neurons. This arrangement allows us to process our sensory environment through detecting the edge orientation of a touched object for instance, and do things like read braille.
These spatially complex receptive fields can studied by using a feedforward neural network to model the tactile periphery. By understanding the processing at the level of the tactile periphery, we can better understand the information that is fed to the CNS for further processing and coordination.
Compressed sampling allows a system to reconstruct a sparse signal with much fewer measurements when the sampling of a sparse input space is sufficiently random. It is our goal to incorporate compressed sampling into a neural network model of the tactile periphery to gain insight into the possible evolutionary processes that resulted in our ability to sense tactile details.