-
Notifications
You must be signed in to change notification settings - Fork 0
Experiment #1
Blake Edwards edited this page Dec 2, 2021
·
3 revisions
Can we create an inheritance algorithm that creates a cell with high fitness at a rate faster than random initialization?
In this experiment, we will be testing a very simple inheritance algorithm to see if we can create an evolutionary process that can beat random cell neural network weight initialization.
If we use the simple neural network inheritance process, the top cell fitness will increase at a rate higher than the top fitness of the randomly initialized cells.
For the control group, during every iteration we will randomly initialize the neural network. We will then let the cell run its course until death and record the fitness. We will also record the top cell fitness over time.