Lifelong Learning with Spiking Neural Networks

Date of Award

Spring 2020

Degree Name

Bachelor of Science

Department

Computer Science & Mathematics; College of Arts & Sciences

First Advisor

Dr. Vitaly Ford

Abstract

A long-standing challenge in machine learning has been the ability of lifelong learning. This supports machines that retain continually learned knowledge and apply it to new and different tasks. In an effort to address this problem, some recent work has moved toward more biologically-inspired computational methods. Methods such as spiking neural networks have proven to be a promising solution, especially when equipped with unsupervised synaptic-timing-dependent plasticity learning rules. Here, we compare existing learning rules through an experiment implemented using the BindsNET Python library, and highlight a new direction towards a more generalized artificial intelligence.

This document is currently not available here.

Share

COinS
 

Lifelong Learning with Spiking Neural Networks

A long-standing challenge in machine learning has been the ability of lifelong learning. This supports machines that retain continually learned knowledge and apply it to new and different tasks. In an effort to address this problem, some recent work has moved toward more biologically-inspired computational methods. Methods such as spiking neural networks have proven to be a promising solution, especially when equipped with unsupervised synaptic-timing-dependent plasticity learning rules. Here, we compare existing learning rules through an experiment implemented using the BindsNET Python library, and highlight a new direction towards a more generalized artificial intelligence.