Let $V$ be the space of all continuous real valued functions on the interval $[1,4]$ with the inner product defined by:

$$\langle f,g\rangle = \int_1^{4} f(t)g(t)\,dt.$$

(i) Find an orthonormal basis of the space $W$ of polynomials of degree less than or equal to $4$.

Here is my attempt:

You should use the [Gram–Schmidt process][1]: Take any three linearly independent polynomials, for example $1,x,x^2$. Now apply the process to this set. $$\|1\|^2=\langle1,4\rangle=\int_1^4 dt=4-1\quad\Rightarrow\quad p_0(x)=\frac{1}{\|4\|}$$ $$proj_{p_0}(v_1)=\langle v_1,p_0 \rangle p_0=p_0\cdot\langle x,p_0 \rangle=\frac{1}{\|4\|}\int_1^4dt$$ and then $u_1=v_1-proj_{p_0}(v_1)$ and $p_1=\frac{u_1}{\|u_1\|}$ and so on.

(ii) Find the polynomial of degree less than or equal to 4 that is closest to $\log t$.

Can you please help me with this? I saw this in a numerical analysis book and was trying to solve this problem the other day. This is what I tried to do but I know this is probably wrong. I have come to the point of exhaustion trying to teach myself this. I tried following other examples but they are not the same. Sorry for not being able to do much.


First of all, you should know how the Gram-Schmidt proccess works (see Wikipedia, for example).

Let $\left\{p_1,\ldots,p_5\right\}$ be a basis for the set $W$ of polynomials (in $[1,4]$) of degree $\leq 4$. For example, you can take $p_i(x)=x^{i-1}$. Apply the Gram-Schmidt process as explained in Wikipedia. You'll obtain 5 polynomials, say $q_1,\ldots,q_5$ s.t. $<q_i,q_j>=\delta_{ij}=\begin{cases}1&\text{, if }i=j\\ 0&\text{ otherwise}\end{cases}$. Remember that, when applying the Gram-Schmidt process, you'll have to calculate some polynomial integrals, which can take quite some time...

The set $\left\{q_1,\ldots,q_5\right\}$ is an orthonormal basis for $W$.

Now, to find polynomial in $W$ that is closest to $\log$, you should already know that this is given by $$p=\sum_{i=1}^5<q_i,\log>q_i$$ that is, the polynomial $p$ is given by $$p(t)=\sum_{i=1}^5\left(\int_1^4q_i(s)\log(s)ds\right)q_i(t)$$ for every $x\in[1,4]$.

  • $\begingroup$ Thanks for helping. I see how this works now. $\endgroup$ – user99744 Nov 3 '13 at 22:30

Your description of the Gram-Schmidt method is somewhat wrong. For instance, if $\|1\|^2=3$, then the corresponding unit polynomial must be $1/\|1\|=1/\sqrt{3}$.

Also, the vector space of polynomials of degree at most 4 is a five-dimensional vector space. You won't find a basis for it consisting of only three polynomials.

What are $v_1$? I don't think you are carrying it out correctly. See here, for example.

For part (ii), it's not just any random "closest" polynomial to $\log x$ that you need to find, you should note that it is closest in the sense of the inner product defined in your question. You are looking for a polynomial belonging to the vector space $W\subset V$ that is closest in the given norm to the given element $\log x$ of the space $V$.

So if $q(x) = \sum_{k=0}^{k=4} c_k u_k(x)$ is that polynomial, then you should know already that $$\langle \log x-q(x), p(x)\rangle=0$$ for every polynomial $p\in W$. In particular $$\langle \log x - q(x), u_k\rangle=0,$$ which gives you the five equations you need to solve for $c_k$. All this follows from the properties of orthonormal bases and inner product vector spaces, and is standard fare in textbooks on real analysis.

  • $\begingroup$ +1 Thanks. I had an idea like this but was not sure so I did it this way. Thank you for explaining it to me. $\endgroup$ – user99744 Nov 3 '13 at 22:31

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.