And how is alpha calculated?
In: 0
Well I don’t know what you mean by alpha but the line search method of finding local minimums or maximums work like this. In general you got a function f(x,y,z,…) for simplicity lets say we have a collection of variables in the imput space and use a vector r. r=(x,y,z,…) so we got f(r) take the derivative or with more than one variable the gradient of the function grad(f(r)). The grad(f) will give you a vector that points towards the largest increase of the function as this vector gets shorter you are approaching a local maximum when taking -grad(f) you can approach a local minimum.
The less general case would be when you have f(x) and you only take the derivative of it at a given point. The derivitive will tell you the slope of the tangent line at that point. If the derivative is positive the function is increasing if negative the function is decreasing. As the slope’s absolute value gets smaller you are approaching a local min or max.
Latest Answers