In the previous chapter the vehicles had what is seemed to be some sort of “knowledge flow”. It’s like the vehicle was doing some sort of calculations before moving. And this can be modified into the concept of “Logic”. Here Braitenberg took the knowledge of what is standard neural network. And activation function. In our vehicle we have two kind of activation.
- Inhibitory
- Exhibitory
We can add them up and create something like the “Activation function RELU”. And like the modern computer with big and sophisticated enough logic, we can calculate/make the vehicle do any sort of calculation and logical work.
But this also begs the question, how can we encorporate memory here?
In computer the idea is, theres computer memory. And the calculations are retrieved from the memory and saved on to the memory. The solution is fairly simple. With large enough network of threshold we can store the data/output of one activation. Consider two thresholds are connected with each other.
For example: A thresholds X are made up of inhibitory and exhibitory activations. And the resulting threshold is in series connection with Y. So, if we revisit the value of Y than we shall always have the logical calculation X made in the past time.
3.12.24 -Abrar Matrikel-Nr: 5012300