Machine Learning as Accelerated Evolution

Each neuron in a neural network can be considered just like a DNA. When some nucleotide changes, we call it a mutation. The result is a different living creature than its ancestors.

We call this “forward propogation” in the Machine Learning literature. We can see the immediate result of that mutation. But is it bad or good for our purposes? Then we calculate the “goodness” of that mutation, and we call it “loss”.

If you only change a small portion of a living animal and then wait for years for it to reproduce, it may take tens of thousands of years to see the effects of that evolution. Why wait if you can see the results immediately?

Use Machine Learning