Great, that's all we need
OK, so the key point is that the differentiation you've learnt is like:
f(x)=log(x)⟹dxdf=x1.
We want a nice consistent way to define differentiation in the case of (for example)
f(x,y)=ylog(x).
There's a very simple way we do this:
∂y∂f=log(x),∂x∂f=xy.
That is, we differentiate as if everything is constant except the variable we're differentiating with respect to. I'll now explain this better :P
Now, the key thing is that, just as a stationary point of
f(x)=x2 can be found by differentiating and setting to 0 - that is,
dxdf=2x=0 so
x=0 is the stationary point - so we can find a stationary point of
f(x,y) by partial-differentiating and setting to 0. Think of it as: we're making a 3D diagram with x and y on the x and y axes, and with height representing f(x,y) at the point (x,y). Then any stationary point (for example, a minimum - that looks like a bowl on our 3D diagram) must be stationary if we slice through the x-axis or the y-axis.
Example series of pictures:
Now, the key thing is that we can easily do this "slicing" just by setting x or y to be constant (in the example, I've set x=0). But if x is constant, then if we differentiate with respect to y, we don't care about x - it just behaves as if it were (say) 2.
So
∂y∂(x2+y2)=2y, because the x^2 was constant so when we differentiated, it became 0.
∂x∂(x2sin(y))=2xsin(y), because y was constant so sin(y) was constant.
OK so far? That's partial differentiation in a nutshell - my next post, if you're with me so far, will be how to solve the triangles question using this.