LMS Algorithm is the algorithm that adjust the value of weights to minimize the mean square of the error. Ultimately, we want to get an output that is close to desired value through LMS algorithm.
So, the structure of this algorithm is like the picture below. You have a set of inputs on the left side, and you have a set of weights next to set of inputs. And the output is going to be the sum of each input multiplied by each weight. With this output, we are going to compare this output,y with desired output,d. And the error is equal to the difference between actual output,y and desired output,d. This is one circulation. And then you will re-set the value of weights and apply input sets again. And you will see what Mean Square Error. AND basically you are doing this over and over again until the error is almost close to zero.
In a word, LMS Algorithm is the algorithm to optimize the setting of weights to take the error to as minimized as it could go.
BUT, LMS algorithm itself has some drawbacks. One of these is error in steady-state and convergence rate are contradictory mutually. In order to overcome this shortcoming, we can apply VSS(Variable Step Size) LMS algorithm to adjust the step size during the process of convergence.