Modern neural network models use a reduced form of the ‘integrate-and-fire’ representation of real neurons in the human brain. However, there are many more complex models of human neurons that rely on differential equations that model the rate-encoded signals that the human brain typically deals with for communicating information. Recently, it’s been found that these models are possibly valuable in the control systems and robotic control domain of artificial intelligence. A question is, how do these neurons work, and how can we use them in AI based control tasks?
Solution:
I implemented various models of spiking neurons in python and demonstrated how they could be chained together for propagating rate limited signals through a spiking neural network.
Methods:
Various neural spiking models implementing things like Hebbian learning.
Frameworks and Platforms:
Python, my own custom implemented code
Outcomes:
Developed a set of code implementing spiking neural networks and demonstrated how they could be chained together to propagate information.
