#ml #voting #ensembling #err #borda #bucklin #condorcet #coombs #reciprocalranking #instantrunoff #fs #featureselection
Гуглил методы голосования (БордА и прочее), и неожиданно наткнулся на их применение в... отборе признаков! Вот уж чего никогда не видел раньше. Есть некий шанс, что это полезно, т.к. один из методов такого "демократического ансамблирования" (Ensemble Reciprocal Ranking) зарулил "лучший одиночный метод" (это был SHAP).
https://towardsdatascience.com/ensemble-feature-selection-for-machine-learning-c0df77b970f9
Гуглил методы голосования (БордА и прочее), и неожиданно наткнулся на их применение в... отборе признаков! Вот уж чего никогда не видел раньше. Есть некий шанс, что это полезно, т.к. один из методов такого "демократического ансамблирования" (Ensemble Reciprocal Ranking) зарулил "лучший одиночный метод" (это был SHAP).
https://towardsdatascience.com/ensemble-feature-selection-for-machine-learning-c0df77b970f9
#voting
"First formulated in Arrow’s doctoral dissertation (published as the monograph Social Choice and Individual Values [Arrow 1951], Arrow’s Impossibility Theory can be stated as follows:
When voters have three or more alternatives, there is no voting method that can convert the ranked preferences of individuals into a community-wide transitive ranking of those alternatives, while also meeting a pre-specified set of fairness conditions in every election.
The study of specific voting methods and their drawbacks actually dates back well before Arrow’s twentieth-century work. Indeed, Iain McLean has remarked that “the theory of voting has in fact been discovered four times and lost three times” [McLean 1990, p. 99]. Arrow, of course, was responsible for the fourth discovery. McLean’s 1990 article examines the first discovery, made at the hands of two medieval thinkers, Ramon Lull (c. 1235–1315) and Nicolas of Cusa (1401–1464), within the context of ecclesiastical elections. More recently, McLean [2019] has written about the third discovery by Charles Dodgson (1832–1898), the British mathematician more widely known as Lewis Carroll, who was motivated to write on the topic as a result of certain election decisions made by the faculty at Christ Church, Oxford. Yet, as those familiar with today’s treatment of voting theory will know, none of the names of Lull, Cusa or Dodgson/Carroll are generally associated with the topic.
In contrast, the second time that this discovery was made involved two late eighteenth-century French mathematicians for whom certain key ideas of voting theory are now named: Jean Charles, Chevalier de Borda (1733–1799) and Marie-Jean-Antoine-Nicolas de Caritat, Marquis de Condorcet (1743–1794)."
https://maa.org/book/export/html/2361819
"First formulated in Arrow’s doctoral dissertation (published as the monograph Social Choice and Individual Values [Arrow 1951], Arrow’s Impossibility Theory can be stated as follows:
When voters have three or more alternatives, there is no voting method that can convert the ranked preferences of individuals into a community-wide transitive ranking of those alternatives, while also meeting a pre-specified set of fairness conditions in every election.
The study of specific voting methods and their drawbacks actually dates back well before Arrow’s twentieth-century work. Indeed, Iain McLean has remarked that “the theory of voting has in fact been discovered four times and lost three times” [McLean 1990, p. 99]. Arrow, of course, was responsible for the fourth discovery. McLean’s 1990 article examines the first discovery, made at the hands of two medieval thinkers, Ramon Lull (c. 1235–1315) and Nicolas of Cusa (1401–1464), within the context of ecclesiastical elections. More recently, McLean [2019] has written about the third discovery by Charles Dodgson (1832–1898), the British mathematician more widely known as Lewis Carroll, who was motivated to write on the topic as a result of certain election decisions made by the faculty at Christ Church, Oxford. Yet, as those familiar with today’s treatment of voting theory will know, none of the names of Lull, Cusa or Dodgson/Carroll are generally associated with the topic.
In contrast, the second time that this discovery was made involved two late eighteenth-century French mathematicians for whom certain key ideas of voting theory are now named: Jean Charles, Chevalier de Borda (1733–1799) and Marie-Jean-Antoine-Nicolas de Caritat, Marquis de Condorcet (1743–1794)."
https://maa.org/book/export/html/2361819