%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 13 0 R 14 0 R 15 0 R ] /Type /Pages /Count 12 >> 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 (2018) /EventType (Poster) /Description-Abstract (We introduce a principled approach for unsupervised structure learning of deep neural networks\056 We propose a new interpretation for depth and inter\055layer connectivity where conditional independencies in the input distribution are encoded hierarchically in the network structure\056 Thus\054 the depth of the network is determined inherently\056 The proposed method casts the problem of neural network structure learning as a problem of Bayesian network structure learning\056 Then\054 instead of directly learning the discriminative structure\054 it learns a generative graph\054 constructs its stochastic inverse\054 and then constructs a discriminative graph\056 We prove that conditional\055dependency relations among the latent variables in the generative graph are preserved in the class\055conditional discriminative graph\056 We demonstrate on image classification benchmarks that the deepest layers \050convolutional and dense\051 of common networks can be replaced by significantly smaller learned structures\054 while maintaining classification accuracy\055\055\055state\055of\055the\055art on tested benchmarks\056 Our structure learning algorithm requires a small computational cost and runs efficiently on a standard desktop CPU\056) /Producer (PyPDF2) /Title (Constructing Deep Neural Networks by Bayesian Network Structure Learning) /Date (2018) /ModDate (D\07220190218220353\05508\04700\047) /Published (2018) /Type (Conference Proceedings) /firstpage (3047) /Book (Advances in Neural Information Processing Systems 31) /Description (Paper accepted and presented at the Neural Information Processing Systems Conference \050http\072\057\057nips\056cc\057\051) /Editors (S\056 Bengio and H\056 Wallach and H\056 Larochelle and K\056 Grauman and N\056 Cesa\055Bianchi and R\056 Garnett) /Author (Raanan Y\056 Rohekar\054 Shami Nisimov\054 Yaniv Gurwicz\054 Guy Koren\054 Gal Novik) /lastpage (3058) >> endobj 3 0 obj << /Type /Catalog /Pages 1 0 R >> endobj 4 0 obj << /Contents 16 0 R /Parent 1 0 R /Resources 17 0 R /MediaBox [ 0 0 612 792 ] /Annots [ 35 0 R 36 0 R 37 0 R 38 0 R 39 0 R 40 0 R 41 0 R 42 0 R 43 0 R 44 0 R 45 0 R 46 0 R 47 0 R 48 0 R 49 0 R 50 0 R 51 0 R 52 0 R 53 0 R 54 0 R 55 0 R 56 0 R 57 0 R 58 0 R 59 0 R 60 0 R 61 0 R 62 0 R 63 0 R 64 0 R 65 0 R 66 0 R ] /Type /Page >> endobj 5 0 obj << /Contents 67 0 R /Parent 1 0 R /Resources 68 0 R /MediaBox [ 0 0 612 792 ] /Annots [ 99 0 R 100 0 R 101 0 R 102 0 R 103 0 R 104 0 R 105 0 R 106 0 R 107 0 R 108 0 R 109 0 R 110 0 R 111 0 R 112 0 R 113 0 R 114 0 R 115 0 R 116 0 R 117 0 R 118 0 R 119 0 R 120 0 R 121 0 R 122 0 R 123 0 R 124 0 R 125 0 R 126 0 R ] /Type /Page >> endobj 6 0 obj << /Contents 127 0 R /Parent 1 0 R /Resources 128 0 R /MediaBox [ 0 0 612 792 ] /Annots [ 161 0 R 162 0 R 163 0 R 164 0 R 165 0 R 166 0 R 167 0 R 168 0 R 169 0 R 170 0 R 171 0 R 172 0 R 173 0 R 174 0 R 175 0 R 176 0 R 177 0 R 178 0 R 179 0 R ] /Type /Page >> endobj 7 0 obj << /Contents 180 0 R /Parent 1 0 R /Resources 181 0 R /MediaBox [ 0 0 612 792 ] /Annots [ 218 0 R 219 0 R 220 0 R 221 0 R 222 0 R ] /Type /Page >> endobj 8 0 obj << /Contents 223 0 R /Parent 1 0 R /Resources 224 0 R /MediaBox [ 0 0 612 792 ] /Annots [ 255 0 R 256 0 R 257 0 R 258 0 R 259 0 R 260 0 R 261 0 R ] /Type /Page >> endobj 9 0 obj << /Contents 262 0 R /Parent 1 0 R /Resources 263 0 R /Group 337 0 R /MediaBox [ 0 0 612 792 ] /Annots [ 361 0 R 362 0 R 363 0 R 364 0 R 365 0 R 366 0 R 367 0 R 368 0 R 369 0 R 370 0 R 371 0 R 372 0 R 373 0 R 374 0 R 375 0 R ] /Type /Page >> endobj 10 0 obj << /Contents 376 0 R /Parent 1 0 R /Resources 377 0 R /MediaBox [ 0 0 612 792 ] /Annots [ 378 0 R 379 0 R 380 0 R 381 0 R 382 0 R 383 0 R 384 0 R 385 0 R 386 0 R 387 0 R 388 0 R 389 0 R 390 0 R 391 0 R 392 0 R 393 0 R 394 0 R 395 0 R 396 0 R 397 0 R 398 0 R 399 0 R 400 0 R 401 0 R 402 0 R 403 0 R 404 0 R 405 0 R 406 0 R 407 0 R 408 0 R 409 0 R 410 0 R 411 0 R 412 0 R 413 0 R 414 0 R 415 0 R 416 0 R 417 0 R 418 0 R 419 0 R 420 0 R 421 0 R 422 0 R 423 0 R ] /Type /Page >> endobj 11 0 obj << /Contents 424 0 R /Parent 1 0 R /Resources 425 0 R /MediaBox [ 0 0 612 792 ] /Annots [ 446 0 R 447 0 R 448 0 R 449 0 R 450 0 R 451 0 R 452 0 R ] /Type /Page >> endobj 12 0 obj << /Contents 453 0 R /Parent 1 0 R /Resources 454 0 R /MediaBox [ 0 0 612 792 ] /Annots [ 459 0 R 460 0 R 461 0 R 462 0 R 463 0 R 464 0 R 465 0 R 466 0 R 467 0 R 468 0 R 469 0 R 470 0 R 471 0 R 472 0 R 473 0 R 474 0 R 475 0 R 476 0 R 477 0 R 478 0 R ] /Type /Page >> endobj 13 0 obj << /Contents 479 0 R /Parent 1 0 R /Type /Page /Resources 480 0 R /MediaBox [ 0 0 612 792 ] >> endobj 14 0 obj << /Contents 481 0 R /Parent 1 0 R /Type /Page /Resources 482 0 R /MediaBox [ 0 0 612 792 ] >> endobj 15 0 obj << /Contents 483 0 R /Parent 1 0 R /Resources 484 0 R /MediaBox [ 0 0 612 792 ] /Annots [ 485 0 R ] /Type /Page >> endobj 16 0 obj << /Length 3046 /Filter /FlateDecode >> stream xڭYKs6W,.A|T.L$l<[[I&%)R^ϯ߯M9r7Boۛ7<ȕ 5EAkͿo|$ R:?x*A*/б~t0S9nteKSo<>oTw-/ҼPjn~{O/ADVAif5gݹM~rrw֠DB7_8tAC[U"BV:i:yK8Ҍeɘ;~pno!D2z4 n^];Z@'DZmQgNGnǍmA8;0NHr^no\V=ԇtE,Yg )n'BWEO љ?or훶fo.?^_%_)v~ vU9,
-LBt3QqFRBs-ȷID:6}m`l2zv:ޫOp
X&sx_6f縈+N|2: ){p#A,XNQ9,)A)YB"$D]3l}@,WL⹃D8,VJ{S۠Cx@\p0Isi{EX:̼8^}H<X1 d]8"ZY{?DQ6 ~m·y