« first day (20 days earlier)    last day (15 days later) » 

5:55 PM
Hi @Semiclassical , I took off last week for Thanksgiving weekend, but I'm at work now :)
 
 
1 hour later…
7:12 PM
Converting our problem into a BIBD:
the set of varieties $X = {v_1, v_2, \ldots, v_{32}}$ are the red nodes
the blocks in $\mathcal B = \{\!\{b_1, b_2, \ldots, b_{28}\}\!\}$ are the adjacencies of the blues nodes 1 through 28
im using that notation for multiset but in this case I think it's a set
so $b_3$ are all the varieties adjacent to red node $v_3$
not sure how many times $r$ each variety appears across all adjacencies, leave it undetermined
each blue vertex allows for 8 adjacencies, so $k=8$
if I'm correct that the 2-multiplicity of length-2 paths between red nodes corresponds to each pair $v_i, v_j: i \ne j$ occurring in exactly 2 blocks, then $\lambda = 2$
by the second equation of block designs, $r(k-1) = \lambda(v-1)$
so $r \cdot 7 = 2 \cdot 31$ which doesn't have an integer solution
so the naive approach, of taking all parameters to be maximal according to the physical constraints, does not lend itself to a BIBD
and actually the parameters are rigid, since $\lambda \cdot (v-1) = 2 \times 31$ is irreducible
or whatever the right word is for can't-be-factored-further
see any mistakes?
so we drop the condition that the design be balanced
a pairwise balanced design PBD is like a BIBD but $k$ need not be constant
or, actually, better idea
seek a BIBD for $v < 32$, and then the $32-v$ left will be dealt with non-optimally
I guess the first one to try would be $v = 29$
 
7:51 PM
R says there is no $(29,8,2)$ BIBD
well if that's the case let's just make a design
 
8:21 PM
forget PBD even, just to see what we come up with
okay its not obvious to me how to get $M$ from the incidence matrix $A'$
oh wait, I only looked for BIBD once, I need to iterate
 
haven't had a chance to look at this for while tbh
 
well nothing happened from then until today
I didnt work last week
 
well, the currient salient problem
how to convert between design matrix $A'$ and biadjacency matrix $M$
I figured out how to convert parameters
 
8:33 PM
I'm trying to deal with 29 vertices because I proved that there isn't a BIBD on 32 vertices that satisfies our constraints, so BIBD algorithms wont find one
unfortunately its easier to prove there isn't a BIBD than to prove there is, so to check 29 vertices I'm using R's algorithms that aren't completely random, but they don't guarantee the non-existence of a solution if the algorithms dont find a solution
the 'blocks' are the sets of adjacencies of each blue vertex
each blue vertex corresponds to a set of red vertices adjacent to it
but the incidence of block $i$ to red vertex $j$ is not the same as the incidence of red vertex $j$ to blue vertex $k$
though it's 'obvious' to me that the data of one can be extracted from the other
somehow
 
not going to be able to look at this any time soon
 
next week or next month or...?
or not at all
 
not today, at any rate :P
 
I have finals in December so im not sure how much ill do in the coming month
 
right
 

« first day (20 days earlier)    last day (15 days later) »