- #1

- 47

- 3

## Main Question or Discussion Point

I had seen a documentary about an algorithm that uses notions of evolution to deduce the equation of motion of a system by sampling a variable connected with the system.

For example, they used the double pendulum case where they sampled the position of the free end of the pendulum and arrived at the equation of motion using curve fitting. For example, if there was a Π/2 in the equation, it would show up as 1.5707.

Here's my vague recollection of how they did it. They used a lot of candidate functions like, x, x^2, exp(x), exp(2x), etc. As you process the samples, some functions better fit the data than others and hence "survive" while the others "die" out. Eventually as the number of samples tends to infinity, the resultant equation tends to the true equation of motion of the system.

The algorithm is very likely in an open source package.

If anyone's heard of something like this, could you please tell me how I could find more info on it.

For example, they used the double pendulum case where they sampled the position of the free end of the pendulum and arrived at the equation of motion using curve fitting. For example, if there was a Π/2 in the equation, it would show up as 1.5707.

Here's my vague recollection of how they did it. They used a lot of candidate functions like, x, x^2, exp(x), exp(2x), etc. As you process the samples, some functions better fit the data than others and hence "survive" while the others "die" out. Eventually as the number of samples tends to infinity, the resultant equation tends to the true equation of motion of the system.

The algorithm is very likely in an open source package.

If anyone's heard of something like this, could you please tell me how I could find more info on it.