BitShares Forum
Other => Graveyard => Follow My Vote => Topic started by: CLains on July 11, 2014, 11:53:46 pm
-
I have not really studied voting theory yet, but found some great links on the topic.
http://en.wikipedia.org/wiki/Voting_system
http://plato.stanford.edu/entries/voting-methods/
http://en.wikipedia.org/wiki/Social_choice_theory
The field seems quite comprehensive.
-
I have not really studied voting theory yet, but found some great links on the topic.
http://en.wikipedia.org/wiki/Voting_system
http://plato.stanford.edu/entries/voting-methods/
http://en.wikipedia.org/wiki/Social_choice_theory
The field seems quite comprehensive.
At the Algorithmic Decision Theory conference (Guerin,
Allen, and Goldsmith, 2013), we presented a heuristic algo-
rithm, earlier proposed by Guerin (2012), for learning CP-
nets from user queries. Our algorithm differs from other ap-
proaches in that: 1. It learns through elicitation rather than
passive observation. 2. It can employ a richer set of queries,
such as outcome comparisons that can differ in any number
of variables and attribute comparisons that offer a heuris-
tic for faster learning. 3. The algorithm is robust: it will al-
ways output a CP-net, never failure. 4. While the algorithm
does not always recover the original CP-net, the learned CP-
net that it outputs is guaranteed to be consistent with the
original on all queries encountered in the learning process.
5. The algorithm is efficient, assuming a bound on in-degree;
that is, if a preference order can be represented as a CP-net,
the algorithm learns a CP-net in time O(n p ), where n is the
number of nodes (features) and p is a bound on the num-
ber of parents a node may have. 6. We employed an exper-
imental approach to evaluate our algorithm, simulating the
query process using randomly generated CP-nets. For the
conference paper, I planned and performed additional ex-
periments, including a series that used statistical sampling
when exhaustive analysis was infeasible. I also modified the
algorithm that randomly generated CP-nets to exclude the
possibility of degenerate CPTs that could conflict with the
dependency graph. The resulting set of experiments demon-
strated that the learned CP-nets agree with the originals on a
high percent of non-training preference comparisons.
http://www.aaai.org/ocs/index.php/AAAI/AAAI14/paper/viewFile/8641/8743