%PDF-1.3 1 0 obj << /Kids [ 4 0 R 5 0 R 6 0 R 7 0 R 8 0 R 9 0 R 10 0 R 11 0 R 12 0 R ] /Type /Pages /Count 9 >> endobj 2 0 obj << /Subject (Neural Information Processing Systems http\072\057\057nips\056cc\057) /Publisher (Curran Associates\054 Inc\056) /Language (en\055US) /Created (2015) /EventType (Poster) /Description-Abstract (Stochastic search algorithms are general black\055box optimizers\056 Due to their ease of use and their generality\054 they have recently also gained a lot of attention in operations research\054 machine learning and policy search\056 Yet\054 these algorithms require a lot of evaluations of the objective\054 scale poorly with the problem dimension\054 are affected by highly noisy objective functions and may converge prematurely\056 To alleviate these problems\054 we introduce a new surrogate\055based stochastic search approach\056 We learn simple\054 quadratic surrogate models of the objective function\056 As the quality of such a quadratic approximation is limited\054 we do not greedily exploit the learned models\056 The algorithm can be misled by an inaccurate optimum introduced by the surrogate\056 Instead\054 we use information theoretic constraints to bound the \140distance\047 between the new and old data distribution while maximizing the objective function\056 Additionally the new method is able to sustain the exploration of the search distribution to avoid premature convergence\056 We compare our method with state of art black\055box optimization methods on standard uni\055modal and multi\055modal optimization functions\054 on simulated planar robot tasks and a complex robot ball throwing task\056The proposed method considerably outperforms the existing approaches\056) /Producer (PyPDF2) /Title (Model\055Based Relative Entropy Stochastic Search) /Date (2015) /ModDate (D\07220160203121719\05508\04700\047) /Published (2015) /Type (Conference Proceedings) /firstpage (3523) /Book (Advances in Neural Information Processing Systems 28) /Description (Paper accepted and presented at the Neural Information Processing Systems Conference \050http\072\057\057nips\056cc\057\051) /Editors (C\056 Cortes and N\056D\056 Lawrence and D\056D\056 Lee and M\056 Sugiyama and R\056 Garnett) /Author (Abbas Abdolmaleki\054 Rudolf Lioutikov\054 Jan R\056 Peters\054 Nuno Lau\054 Luis Pualo Reis\054 Gerhard Neumann) /lastpage (3531) >> endobj 3 0 obj << /Type /Catalog /Pages 1 0 R >> endobj 4 0 obj << /Contents 13 0 R /Parent 1 0 R /Type /Page /Resources 14 0 R /MediaBox [ 0 0 612 792 ] >> endobj 5 0 obj << /Contents 40 0 R /Parent 1 0 R /Type /Page /Resources 41 0 R /MediaBox [ 0 0 612 792 ] >> endobj 6 0 obj << /Contents 62 0 R /Parent 1 0 R /Type /Page /Resources 63 0 R /MediaBox [ 0 0 612 792 ] >> endobj 7 0 obj << /Contents 80 0 R /Parent 1 0 R /Type /Page /Resources 81 0 R /MediaBox [ 0 0 612 792 ] >> endobj 8 0 obj << /Contents 90 0 R /Parent 1 0 R /Resources 91 0 R /Group 229 0 R /MediaBox [ 0 0 612 792 ] /Type /Page >> endobj 9 0 obj << /Contents 255 0 R /Parent 1 0 R /Type /Page /Resources 256 0 R /MediaBox [ 0 0 612 792 ] >> endobj 10 0 obj << /Contents 257 0 R /Parent 1 0 R /Resources 258 0 R /Group 361 0 R /MediaBox [ 0 0 612 792 ] /Type /Page >> endobj 11 0 obj << /Contents 363 0 R /Parent 1 0 R /Resources 364 0 R /Group 409 0 R /MediaBox [ 0 0 612 792 ] /Type /Page >> endobj 12 0 obj << /Contents 1001 0 R /Parent 1 0 R /Type /Page /Resources 1002 0 R /MediaBox [ 0 0 612 792 ] >> endobj 13 0 obj << /Length 2941 /Filter /FlateDecode >> stream xڵks۸5cHK]8kLfYIB߯> kz3" bߠB/'Ћ(,,2@(3Oezvxy{Op ĻyQ<2n+T^V8 ZzUTȀl֮$~śۧaKU~ zw3,H$"8f6(F,զ2u#kuQʽ,(sQ"s"ȢOG3,as*q {&A{뙘.#\nkmA߯7l3>yE6n,+}[cyzxй2~AX|q~(щ_?uDA\ o e{Pu!T-H$AT&$}zb?.}ҧ'ʖU]|Tc#Q&@"˛Zvbni@ -?,͛U)++çV3Jꃒ`};s;_L7w+P}MɃzA*IKV7o$G8e(,W$_NH{~ҠHSSՍ=Zȫ tײkAV\|t.XX?oVZ{m`t>wlx@W힑$S&'PS<|i~