Method of Least square

67 Views Asked by At

Sometimes, the method of least squares can be used for problems that, in principle, could not be addressed by the method.

For example, let's suppose we want to find $a$ and $b$ in order to fit the function $\frac{1}{{a + bx}}$ to the points $(x_i, y_i)$ by minimizing $$min\frac{1}{2}\sum_{i=1}^m (\frac{1}{a + bx_i} - y_i)^2$$

This minimization is not in a form that the method of least squares can solve. However, an idea is to find a transformation $T$ such that $$min\frac{1}{2}\sum_{i=1}^m [T(\frac{1}{a + bx_i}) - T(y_i)]^2$$ is a problem of least squares. In the given case, if we use $T(x) = \frac{1}{x}$, we will have the minimization of $$min\frac{1}{2}\sum_{i=1}^m (a + bx_i - \frac{1}{y_i})^2$$

where we can apply the method of least squares.

Find appropriate transformations for the following functions:

a) $f(x) = ae^{bx}$

b) $f(x) = \frac{1}{{a + bx + cx^2}}$

So, for letter A, I used the transformation $T(x) = \ln(x)$, which provides the desired result. However, in letter B, I'm uncertain about the next steps. How can we transform a quadratic function into a linear function so that we can apply the least squares method?