We research an extension of the usual two-party communication mannequin by which Alice and Bob maintain likelihood distributions $p<\/annotation><\/semantics><\/math><\/span>p<\/span><\/span><\/span><\/span> and q<\/mi><\/mrow>q<\/annotation><\/semantics><\/math><\/span>q<\/span><\/span><\/span><\/span> over domains X<\/mi><\/mrow>X<\/annotation><\/semantics><\/math><\/span>X<\/span><\/span><\/span><\/span> and Y<\/mi><\/mrow>Y<\/annotation><\/semantics><\/math><\/span>Y<\/span><\/span><\/span><\/span>, respectively. Their purpose is to estimate<\/p>\nE<\/mi>x<\/mi>\u223c<\/mo>p<\/mi>,<\/mo>y<\/mi>\u223c<\/mo>q<\/mi><\/mrow><\/msub>[<\/mo>f<\/mi>(<\/mo>x<\/mi>,<\/mo>y<\/mi>)<\/mo>]<\/mo><\/mrow>mathbb{E}_{x sim p, y sim q}[f(x, y)]<\/annotation><\/semantics><\/math><\/span>E<\/span>x<\/span>\u223c<\/span>p<\/span>,<\/span>y<\/span>\u223c<\/span>q<\/span><\/span><\/span><\/span><\/span>\u200b<\/span><\/span><\/span><\/span><\/span><\/span><\/span>[<\/span>f<\/span>(<\/span>x<\/span>,<\/span>y<\/span>)]<\/span><\/span><\/span><\/span><\/p>\nto inside additive error \u03b5<\/mi><\/mrow>varepsilon<\/annotation><\/semantics><\/math><\/span>\u03b5<\/span><\/span><\/span><\/span> for a bounded perform f<\/mi><\/mrow>f<\/annotation><\/semantics><\/math><\/span>f<\/span><\/span><\/span><\/span>, identified to each events. We seek advice from this because the distributed estimation downside. Particular circumstances of this downside come up in quite a lot of areas together with sketching, databases and studying. Our purpose is to grasp how the required communication scales with the communication complexity of f<\/mi><\/mrow>f<\/annotation><\/semantics><\/math><\/span>f<\/span><\/span><\/span><\/span> and the error parameter \u03b5<\/mi><\/mrow>varepsilon<\/annotation><\/semantics><\/math><\/span>\u03b5<\/span><\/span><\/span><\/span>.<\/p>\nThe random sampling method \u2014 estimating the imply by averaging over O<\/mi>(<\/mo>1<\/mn>\/<\/mi>\u03b5<\/mi>2<\/mn><\/msup>)<\/mo><\/mrow>O(1\/varepsilon^2)<\/annotation><\/semantics><\/math><\/span>O<\/span>(<\/span>1\/<\/span>\u03b5<\/span>2<\/span><\/span><\/span><\/span><\/span><\/span><\/span><\/span>)<\/span><\/span><\/span><\/span> random samples \u2014 requires O<\/mi>(<\/mo>R<\/mi>(<\/mo>f<\/mi>)<\/mo>\/<\/mi>\u03b5<\/mi>2<\/mn><\/msup>)<\/mo><\/mrow>O(R(f)\/varepsilon^2)<\/annotation><\/semantics><\/math><\/span>O<\/span>(<\/span>R<\/span>(<\/span>f<\/span>)<\/span>\/<\/span>\u03b5<\/span>2<\/span><\/span><\/span><\/span><\/span><\/span><\/span><\/span>)<\/span><\/span><\/span><\/span> complete communication, the place R<\/mi>(<\/mo>f<\/mi>)<\/mo><\/mrow>R(f)<\/annotation><\/semantics><\/math><\/span>R<\/span>(<\/span>f<\/span>)<\/span><\/span><\/span><\/span> is the randomized communication complexity of f<\/mi><\/mrow>f<\/annotation><\/semantics><\/math><\/span>f<\/span><\/span><\/span><\/span>. We design a brand new debiasing protocol which improves the dependence on 1<\/mn>\/<\/mi>\u03b5<\/mi><\/mrow>1\/varepsilon<\/annotation><\/semantics><\/math><\/span>1\/<\/span>\u03b5<\/span><\/span><\/span><\/span> to be linear as an alternative of quadratic. Moreover we present higher higher bounds for a number of particular courses of capabilities, together with the Equality and Higher-than capabilities. We introduce decrease sure strategies primarily based on spectral strategies and discrepancy, and present the optimality of lots of our protocols: the debiasing protocol is tight for common capabilities, and that our protocols for the equality and greater-than capabilities are additionally optimum. Moreover, we present that amongst full-rank Boolean capabilities, Equality is basically the best.<\/p>\n\n\u2020 College of California, Los Angeles<\/li>\n\u2021 College of California, Berkeley<\/li>\n\u00a7 Institute for Superior Research (IAS)<\/li>\n<\/ul>\n<\/div>\n\n","protected":false},"excerpt":{"rendered":"We research an extension of the usual two-party communication mannequin by which Alice and Bob maintain likelihood distributions ppp and qqq over domains XXX and YYY, respectively. Their purpose is to estimate Ex\u223cp,y\u223cq[f(x,y)]mathbb{E}_{x sim p, y sim q}[f(x, y)]Ex\u223cp,y\u223cq\u200b[f(x,y)] to inside additive error \u03b5varepsilon\u03b5 for a bounded perform fff, identified to each events. We seek […]<\/p>\n","protected":false},"author":2,"featured_media":10205,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[55],"tags":[2583,2158,741,3688],"class_list":["post-10203","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-machine-learning","tag-communication","tag-complexity","tag-distributed","tag-estimation"],"_links":{"self":[{"href":"https:\/\/techtrendfeed.com\/index.php?rest_route=\/wp\/v2\/posts\/10203","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/techtrendfeed.com\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/techtrendfeed.com\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/techtrendfeed.com\/index.php?rest_route=\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/techtrendfeed.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=10203"}],"version-history":[{"count":1,"href":"https:\/\/techtrendfeed.com\/index.php?rest_route=\/wp\/v2\/posts\/10203\/revisions"}],"predecessor-version":[{"id":10204,"href":"https:\/\/techtrendfeed.com\/index.php?rest_route=\/wp\/v2\/posts\/10203\/revisions\/10204"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/techtrendfeed.com\/index.php?rest_route=\/wp\/v2\/media\/10205"}],"wp:attachment":[{"href":"https:\/\/techtrendfeed.com\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=10203"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/techtrendfeed.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=10203"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/techtrendfeed.com\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=10203"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}$