<div dir="auto">It must be the 40 degree Celsius heat wave frying my brain: D(X) is supposed to be the candidates defeating X, not defeated by X. And so D'(X) should be the set of candidates that do not defeat X.<div dir="auto"><br></div><div dir="auto">I apologize... right mental picture ... wrong verbal description!</div><div dir="auto"><br></div><div dir="auto">I have made this same error before ... in the context of Decloned Copeland.</div><div dir="auto"><br></div><div dir="auto">So let me run this past you ...</div><div dir="auto"><br></div><div dir="auto">Let FX) be the First place preference total of all candidates that do not defeat X. If no candidate defeated X, then F(X) would just be the total number of ballots ... which suggests electing the candidate X that maximizes F(X).</div><div dir="auto"><br></div><div dir="auto">Anything wrong with that?</div><div dir="auto"><br></div><div dir="auto">If not, then it is (arguably) the best UD Landau efficient method we could offer for public proposal (un my not overly humble opinion).</div><div dir="auto"><br></div><div dir="auto">Two more days of high temperatures!</div><div dir="auto"><br></div><div dir="auto">-Forest</div><div dir="auto"><br></div><div dir="auto"><br></div></div><br><div class="gmail_quote"><div dir="ltr" class="gmail_attr">El vie., 29 de jul. de 2022 8:30 p. m., Forest Simmons <<a href="mailto:forest.simmons21@gmail.com">forest.simmons21@gmail.com</a>> escribió:<br></div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div dir="auto"><div dir="auto"><p dir="ltr">Kristofer,</p><p dir="ltr">[For the record I am cc-ing this to the EM list.]</p><p dir="ltr">While pondering the geometry of our agenda deas for Landau efficient methods, the following idea for a Universal Domain Landau Efficient method came into focus:<br></p></div><div dir="auto"><br></div><div dir="auto">For each alternative X let D(X) be the set of alternatives that defeat X pairwise, and let D'(X) be its complement. Note that X is in D'(X).</div><div dir="auto"><br></div><div dir="auto">Our idea is that the closer D(X) is to being engulfed by D'X), the closer X is to being undefeated.</div><div dir="auto"><br></div><div dir="auto">So how do we measure how much D'(X) would have to be enlarged in order to swallow up D(X)?</div><div dir="auto"><br></div><div dir="auto">By the distance from the member of D(X) that is furthest from the unswollen D'(X).</div><div dir="auto"><br></div><div dir="auto">The distance of an alternative Y to the set D'(X) is defined as</div><div dir="auto">dist(Y, D'(X))=Min over Z in D'(X) of dist(Y,Z), i.e. the distance from Y to its nearest neighbor Z in D'(X).</div><div dir="auto"><br></div><div dir="auto">So to engulf D(X) the frontier of D'(X) has to be extended outward a distance of Extend(X)=Max over Y in D(X) of dist(Y,D'(X)).</div><div dir="auto"><br></div><div dir="auto">The less extension needed, the better, so we want to elect argmin(Extend(X))</div><div dir="auto"><br></div><div dir="auto">The only thing left is to pick an appropriate distance metric.</div><div dir="auto"><br></div><div dir="auto">If we want to remain in the Universal Domain category, it seems to me that the best gauge of dist(Y, Z) is the disappointment incurred by moving from alternative Y to Z, in other words, the number of ballots that prefer Y over Z.</div><div dir="auto"><br></div><div dir="auto">Note that if A covers B, then it is (in general) easier for D(A) to engulf D'(A) than for D(B) to engulf D'(B), so this method naturally elects uncovered candidates.</div><div dir="auto"><br></div><div dir="auto">There is a natural way to break ties that makes this Landau property snug ... by expanding/extending the engulfing frontier gradually, while keeping track of how much expansion is needed to swallow up the respective candidates one by one ... but for now we won't worry about the details.</div><div dir="auto"><br></div><div dir="auto">Does this make sense?</div><div dir="auto"><br></div><div dir="auto">Any obvious defects?</div><div dir="auto"><br></div><div dir="auto">The basic idea that came from pondering our agenda covering Landau method (in connection with its Banks efficient mimic) is that it minimizes the max distance to D'(X) or "theall(X) union {X}" instead of minimizing the max distance to X itself as per standard MinMax. Similarly, the Banks version minimizes the max distance to a dense subset of D'(X) [a chain that covers D'(X].</div><div dir="auto"><br></div><div dir="auto">["dense" because every member of D'(X) is just one small step away from the chain]</div><div dir="auto"><br></div><div dir="auto">So the topology of digraphs guides our intuition as always!</div><div dir="auto"><br></div><div dir="auto">-Forest</div><div dir="auto"><br></div><div dir="auto"><br></div></div>
</blockquote></div>