Machine Learning Math Experiments – a(n+1)/a(n) = 2 Where a = model( np.random.rand() )

Given c. A problem of finding a and b in c = a + b. Seems impossible when you only know c. From this I wonder if nature has solved problems like this by limiting probability.

Then coming up with a problem of your own. Like

a = model(np.random.rand())
a(n+1)/a(n) = 2 # some constant

Find a model() or function that solves this.

I did find a model() but it was perhaps to easy when it had some memory. One thing was odd. I had to use .5 as a 2 replacement and then 1/model(). From this I guess that a lot of machine learning are technical problems.

Which gives me an idea.

Why not use a try: except in the loop?

 

Machine Learning Physics Idea – Potential, Kinetic And Function Energy? Function Energy is energy that the network could use to solve problems from incoming particles

I wrote an idea before about acceleration chemistry. I think I now have an understanding or a better guess.

From machine learning. When a piece of data comes that would cause too many weights to be recalculated. A ?better strategy would be to let the data pass. That is. Do nothing with it. Get it out of the system as fast as possible.

I think this is what happens in ?acceleration chemistry. If you activate Van der Waals forces between two metals. That holding together function has some purpose. So if an electron with some higher energy comes along. That would mean a function re-calculation. So it could be better to pass the electron by with some acceleration.

So Function Energy is energy that the network could use to solve incoming particles.

So in an electric Vehicle. Perhaps the electric network could be as efficient as slingshot gravity.

The idea is that the electric network takes energy to solve both the Van der Waals function and the particle problem and that there probably is a way to take advantage of this.