Intuition behind the ILATE rule

Often I have wondered about this question, but today I had a chance to recollect it and hence I am posting it here. During high-school days one generally learns Integration and I still loving doing problems on Integration. But I have never understood this idea of integration by parts. Why do we adopt the ILATE or LIATE rule to do problems and what was the reason behind thinking of such a rule?

• Can you elaborate what ILATE or LIATE is? By the way, isn't integration by parts just the analogue of the product rule? – Jonas Teuwen Nov 13 '10 at 17:31
• – anonymous Nov 13 '10 at 17:32
• @Jonas T: L=logarthmic function, A=algebraic functions T=trignometric function, E=exponential function. I for... – anonymous Nov 13 '10 at 17:33
• @Joas T: No jonas, i am talking about the choice of functions $u$ and $v$ when do we integration by parts. – anonymous Nov 13 '10 at 17:35
• I'd personally consider this ILATE thing (I haven't encountered this until now as well) as more of a suggestion than a rule. I don't know... most of the time I've had to use integration by parts, the splitting looks nearly transparent. On the other hand, I've had a hard time explaining to other people why I split the way I do. – J. M. isn't a mathematician Nov 14 '10 at 1:01

The way I see it, when you differentiate an inverse trigonometric function, you don't get another inverse trigonometric function. Instead you get "simpler" functions like $1/(1 + x^2)$ or $1/\sqrt{1-x^2}$. This does not typically happen with the antiderivative of such functions.
So, when using integration by parts $\int u dv = uv - \int v du$, it makes sense to select the inverse trigonometric or logarithmic function to be the one that is the $u$ term.
As a technique for explicitly integrating functions given by formulas as usually seen in calculus classes, integration by parts works because $u'v$ can be easier to integrate than $uv'$. There are multiple ways an integrand can be considered as a product of the form $u'v$, but some choices lead to $uv'$ being of no use, perhaps even harder to integrate. The ILATE mnemonic you mention (or whatever it is) gives some rules that approximate the best guesses as to what will work well. Better than this mnemonic (which I've never thought about) is to just have enough experience doing integrals to be able to see what will work and why (at least for the type of integrals one typically sees in a calculus class).