[Edit: read the bottom first:)]
It is beginning to get too complicated. You can try to make something like this. Denote by H(N) the function that gives you the partitions of something of the shape
where the main block is of length N, and there is an extra "ear" sticking out of size 1x1. Then we can compute the functions G and H simultaneously:
1 2
|
G(N)= 2G(N-1)+G(N-2)+2H(N-2)
H(N)=G(N)+H(N-1)
|
Motivation for the first relation. If your block starts with : or | you get the summand
2G(N-1)
. Otherwise your block must start with a horizontal slab _ or -. The option that you have two horizontal slabs = is accounted by
G(N-2)
. The other two options are to have something like
1 2 3 4 5
|
________
| |
|_______|
| |
|___|
|
or with the longer slab on the bottom. Those are accounted for by the summand
2H(N-2)
.
Motivation for the second relation. There are only two options for the "ear": either it is taken by a one by one block, which accounts for the
G(N-1)
summand, or it is taken by a 2x1 horizontal block, which accounts for the second summand
H(N-1)
.
[Edit:] I just figured out that when I plug in my function H(N-2) back into the first I get exactly your algebgraic relation! It is, in fact, the answer! Now it is clear how to end up this sum: H(0) is well defined, it is the number of ways to split a 1x1 block, which is exactly 1. Therefore, your sum should end like this:
1 2 3
|
G(N) = 2G(N-1)+3G(N-2)+2G(N-3)+2G(N-4)+2G(N-5)+...+2G(1)+2H(0)=
= 2G(N-1)+3G(N-2)+2G(N-3)+2G(N-4)+2G(N-5)+...+2G(1)+2G(0)
|
since we already decided to define G(0)=H(0) to be 1 >:)