Approximate $F$ by a function of the type $f(t) = ae^{bt}$ by the least squares method

F(t) - 132 65 31 17 8

t - ----1---2 -3--4 - 5

This is what my teacher did (sketches because I don't know all the logical steps that he did): $$\min ||F(t)-f(t) || = \min_{a,b} \sum_{i=1}^5 (F(t_i)-ae^{bt_i})^2$$

$$F -ae^{bt}\implies \ln F - \ln ae^{bt} = \ln a + bt $$

$$\min\sum(\ln F(t_i)-\ln a + bt_i)^2$$ $$ \begin{bmatrix} \langle 1,1\rangle & \langle 1,t\rangle \\ \langle t,1\rangle & \langle t,t\rangle \end{bmatrix}\begin{bmatrix} \alpha \\ \beta \end{bmatrix} = \begin{bmatrix} 5 & 15\\ 15 & 55 \end{bmatrix} \begin{bmatrix} \alpha \\ \beta \end{bmatrix} = \begin{bmatrix} \langle \ln F, 1\rangle \\ \langle \ln F,t\rangle \end{bmatrix} $$

I think that that matrix of inner products came from this:

We're trying to find $\min ||F(t) - f(t)||$. This is the same as projecting $F(t)-f(t)$ onto the subspace generated by $f(t)$, so we should have:

$$\langle \ln F-\ln a - bt,1\rangle=0\\ \langle \ln F-\ln a - bt,t\rangle=0$$ which implies

$$\langle \ln F,1\rangle = \langle\ln a,1\rangle + \langle bt,1\rangle\\\langle \ln F,t\rangle = \langle\ln a,t\rangle + \langle bt,t\rangle$$

which is the same as

$$\langle \ln F,1\rangle = \ln a\langle 1,1\rangle + b\langle t,1\rangle\\\langle \ln F,t\rangle = \ln a\langle1,t\rangle + b\langle t,t\rangle$$ which is the system we have above.

Now the question is: which inner product he used to get that matrix? I don't have idea on how to take inner product of those things.


In linear least squares, you are using the little 'el two' $l^2$ inner product since you have discrete data points.

$$<f,g> = \sum_{i=1}^{5} f_ig_i.$$


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.