%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) /Language (en\055US) /Created (2010) /Description-Abstract (Probabilistic models of natural images are usually evaluated by measuring performance on rather indirect tasks\054 such as denoising and inpainting\056 A more direct way to evaluate a generative model is to draw samples from it and to check whether statistical properties of the samples match the statistics of natural images\056 This method is seldom used with high\055resolution images\054 because current models produce samples that are very different from natural images\054 as assessed by even simple visual inspection\056 We investigate the reasons for this failure and we show that by augmenting existing models so that there are two sets of latent variables\054 one set modelling pixel intensities and the other set modelling image\055specific pixel covariances\054 we are able to generate high\055resolution images that look much more realistic than before\056 The overall model can be interpreted as a gated MRF where both pair\055wise dependencies and mean intensities of pixels are modulated by the states of latent variables\056 Finally\054 we confirm that if we disallow weight\055sharing between receptive fields that overlap each other\054 the gated MRF learns more efficient internal representations\054 as demonstrated in several recognition tasks\056) /Producer (Python PDF Library \055 http\072\057\057pybrary\056net\057pyPdf\057) /Title (Generating more realistic images using gated MRF\047s) /Date (2010) /Type (Conference Proceedings) /firstpage (2002) /Book (Advances in Neural Information Processing Systems 23) /Description (Paper accepted and presented at the Neural Information Processing Systems Conference \050http\072\057\057nips\056cc\057\051) /Editors (J\056D\056 Lafferty and C\056K\056I\056 Williams and J\056 Shawe\055Taylor and R\056S\056 Zemel and A\056 Culotta) /Author (Marc\047aurelio Ranzato\054 Volodymyr Mnih\054 Geoffrey E\056 Hinton) /lastpage (2010) >> 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 36 0 R /Parent 1 0 R /Type /Page /Resources 37 0 R /MediaBox [ 0 0 612 792 ] >> endobj 6 0 obj << /Contents 38 0 R /Parent 1 0 R /Type /Page /Resources 39 0 R /MediaBox [ 0 0 612 792 ] >> endobj 7 0 obj << /Contents 74 0 R /Parent 1 0 R /Type /Page /Resources 75 0 R /MediaBox [ 0 0 612 792 ] >> endobj 8 0 obj << /Contents 86 0 R /Parent 1 0 R /Type /Page /Resources 87 0 R /MediaBox [ 0 0 612 792 ] >> endobj 9 0 obj << /Contents 104 0 R /Parent 1 0 R /Type /Page /Resources 105 0 R /MediaBox [ 0 0 612 792 ] >> endobj 10 0 obj << /Contents 108 0 R /Parent 1 0 R /Type /Page /Resources 109 0 R /MediaBox [ 0 0 612 792 ] >> endobj 11 0 obj << /Contents 114 0 R /Parent 1 0 R /Type /Page /Resources 115 0 R /MediaBox [ 0 0 612 792 ] >> endobj 12 0 obj << /Contents 124 0 R /Parent 1 0 R /Type /Page /Resources 125 0 R /MediaBox [ 0 0 612 792 ] >> endobj 13 0 obj << /Length 3085 /Filter /FlateDecode >> stream xڍYY~_7SU#Ļ\vv'd òק/q" 4;|~C _ ϝȽ"y@bTNy9>ß>Ed^$tt T9g^}wmUa4%L^iQW˧vQ?y{Aɟ_|+m|v.RxEt8D^|MGojNʼn]nWwյr֜h8t#3fڱkZpdu¼p#)DgՏnG^9O}FFNy)ܜ埭م.H\fr{;8+Wڧrt51G =KxZ]y^vbrF>Ip6C/H3^ga{UNU=> R}{y$sX5 menƩW5c[gLäzWk~QWDaFp===(](T}.ț`'l21PoTm*OLU+vmey<'FE~`:X\~^x\u|TşWq13fW '^7\krFɉvIo7c43gBl羃 y7@&tX8ڨb9 sB"+%姓! ȟF !)!pO9ӾCWOA%¹#8Rb)Y8P5,[ :I"x+*sg?jHZĒǤxʳL֑ĹAi32 3b_+@G7F6ݵwj$SO, Eé_--B)|1m.;jXW^+ɈeGZ[7PmkkPȈtl~'< Gΐ@C3[cJ\P]߾[mpl8[24I#P w>)Ą`ʌpu1!74r^vPT˭M'P>do(^ ?ZZF_ 8n\H6{BkM]+.*}-jn-!O0.LQ$"Cvm/l~S-.I#s9zcF1&6|2-HKE!ͳˮEe#ck u-NċI$ؒ/ZhC>FM$ keՙZ0GvϢCV={H#oGoWIJvIZrGhк"_Mi8K aZ+?/56M<%^ fAx <H(?.`m &85a+7ISBhl