%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) /Description-Abstract (Bayesian nonparametric hidden Markov models are typically learned via fixed truncations of the infinite state space or local Monte Carlo proposals that make small changes to the state space\056 We develop an inference algorithm for the sticky hierarchical Dirichlet process hidden Markov model that scales to big datasets by processing a few sequences at a time yet allows rapid adaptation of the state space cardinality\056 Unlike previous point\055estimate methods\054 our novel variational bound penalizes redundant or irrelevant states and thus enables optimization of the state space\056 Our birth proposals use observed data statistics to create useful new states that escape local optima\056 Merge and delete proposals remove ineffective states to yield simpler models with more affordable future computations\056 Experiments on speaker diarization\054 motion capture\054 and epigenetic chromatin datasets discover models that are more compact\054 more interpretable\054 and better aligned to ground truth segmentations than competitors\056 We have released an open\055source Python implementation which can parallelize local inference steps across sequences\056) /Producer (PyPDF2) /Title (Scalable Adaptation of State Complexity for Nonparametric Hidden Markov Models) /Date (2015) /ModDate (D\07220160216125343\05508\04700\047) /Published (2015) /Type (Conference Proceedings) /firstpage (1198) /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) /EventType (Poster) /Author (Michael C\056 Hughes\054 William T\056 Stephenson\054 Erik Sudderth) /lastpage (1206) >> endobj 3 0 obj << /Type /Catalog /Pages 1 0 R >> endobj 4 0 obj << /Parent 1 0 R /Rotate 0 /Contents 13 0 R /Resources << /ExtGState 14 0 R /ProcSet [ /PDF /ImageB /Text ] /Font 16 0 R >> /MediaBox [ 0 0 612 792 ] /Annots [ 31 0 R 32 0 R 33 0 R 34 0 R 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 ] /Type /Page >> endobj 5 0 obj << /Parent 1 0 R /Rotate 0 /Contents 47 0 R /Resources << /ExtGState 48 0 R /ColorSpace 49 0 R /XObject 55 0 R /ProcSet [ /PDF /ImageC /Text ] /Font 85 0 R >> /MediaBox [ 0 0 612 792 ] /Annots [ 137 0 R 138 0 R 139 0 R 140 0 R 141 0 R 142 0 R ] /Type /Page >> endobj 6 0 obj << /Parent 1 0 R /Rotate 0 /Contents 143 0 R /Resources << /ExtGState 144 0 R /ColorSpace 145 0 R /XObject 146 0 R /ProcSet [ /PDF /ImageB /ImageC /Text ] /Font 154 0 R >> /MediaBox [ 0 0 612 792 ] /Annots [ 260 0 R 261 0 R 262 0 R 263 0 R 264 0 R 265 0 R 266 0 R ] /Type /Page >> endobj 7 0 obj << /Parent 1 0 R /Rotate 0 /Contents 267 0 R /Resources << /ExtGState 268 0 R /ProcSet [ /PDF /ImageB /Text ] /Font 269 0 R >> /MediaBox [ 0 0 612 792 ] /Annots [ 270 0 R 271 0 R 272 0 R 273 0 R 274 0 R 275 0 R 276 0 R 277 0 R 278 0 R 279 0 R 280 0 R 281 0 R ] /Type /Page >> endobj 8 0 obj << /Parent 1 0 R /Rotate 0 /Contents 282 0 R /Resources << /ExtGState 283 0 R /ProcSet [ /PDF /ImageB /Text ] /Font 284 0 R >> /MediaBox [ 0 0 612 792 ] /Annots [ 285 0 R 286 0 R 287 0 R 288 0 R 289 0 R 290 0 R 291 0 R 292 0 R 293 0 R 294 0 R ] /Type /Page >> endobj 9 0 obj << /Parent 1 0 R /Rotate 0 /Contents 295 0 R /Resources << /ExtGState 296 0 R /XObject 297 0 R /ProcSet [ /PDF /ImageC /Text ] /Font 309 0 R >> /MediaBox [ 0 0 612 792 ] /Annots [ 311 0 R 312 0 R 313 0 R 314 0 R 315 0 R ] /Type /Page >> endobj 10 0 obj << /Parent 1 0 R /Rotate 0 /Contents 316 0 R /Resources << /ExtGState 317 0 R /ProcSet [ /PDF /Text ] /Font 318 0 R >> /MediaBox [ 0 0 612 792 ] /Annots [ 328 0 R 329 0 R 330 0 R 331 0 R 332 0 R 333 0 R 334 0 R 335 0 R 336 0 R 337 0 R 338 0 R 339 0 R 340 0 R 341 0 R 342 0 R 343 0 R 344 0 R 345 0 R ] /Type /Page >> endobj 11 0 obj << /Parent 1 0 R /Rotate 0 /Contents 346 0 R /Resources << /ExtGState 347 0 R /ColorSpace 348 0 R /XObject 349 0 R /ProcSet [ /PDF /ImageB /ImageC /Text ] /Font 378 0 R >> /MediaBox [ 0 0 612 792 ] /Annots [ 389 0 R 390 0 R 391 0 R 392 0 R ] /Type /Page >> endobj 12 0 obj << /Parent 1 0 R /Rotate 0 /Contents 393 0 R /Resources << /ExtGState 394 0 R /ProcSet [ /PDF /Text ] /Font 395 0 R >> /MediaBox [ 0 0 612 792 ] /Type /Page >> endobj 13 0 obj << /Length 7407 /Filter /FlateDecode >> stream x]Is7'Ra)Чnmu{nM&(H[L ZpYW5/7ϛy/ޟ
ɓl^<~#N|{6:y|4ODh,xy:pgXū0ZaWƸI͋'?l}z&&YgrB=ǻvի^03.]O-Kh@AӳrJ(:ei=^.V+}}[)\coJoIOԤkW:fxۏ52؛4SWIEB[j?ljXU89nrxa͚}-qvWIّy+Aྴۯ㴶eif)Ǫ-";VX2%s,;5PZk+w
ʠFv0愭4̹i֛VTO@1oOgxaY`W."ͩ@J)-0FN4~ T=O<[ o/ȸmzrr1j{sB,,%,3â -H=nPWþ*wdKW7uzÓF,4nޔ~i*,Woj6bqxL |¿>͏e.3*|;2"^YPweBa7[0)Pt~ "b&"+sJy]Xi~Z8
$QYBk02)Wd&0"X)
Hui1Ϥ۲+ӻبqQֈTE
Ջ=Q6d
ݜ
; Phi3o};vTR/fFsI
%DV'`p VoTC
y|'!aE'nLt|b/P8H2v.;)˥f^{ ҲhYaЙ SΆpC|>\~( V
?0˟]k{.W%7t|^[-#Lz枝=;}vɦ˷]ҋL7\cQ[q|{."!ٓP$ @G ZxLAנ0.bd:[iQrP#>_3uo?x}IXE:HX/;\7T\ L6P!pzJM,U6BFjŻ,of%IfpW>`WxMsEM4.Yc~
: @ĻW,w 6G)"\>ʮ":KC}iੇ,<",[IMݴ>`/%2ݨ7K@!t
|5^56,ĐY [j d)L+BiqLG~В7)r_!J?WK -)P3u!B1RwjF%rHnU-w/``t0[+BU0[kύ5lQa )qJUzZ9~f$ۿuɦnO9dӢ]%a 9b7SЖFwݑm 6Vzi^^"5KyEln~U^x{H<ڻW2/6NЊro66kxl^%O4UEjIx{<3NEϊ/ ۪uRZ>)yHɿRּX9W
n+=]N 823բIN- AAQЯkTx7q+%<rEa!.N K
A$A*wD삕~2!9oÈyIZIGG6L>GQK2 JSp.,[s2iAT$ߓYyQպ -߮/KbDy݅G9")E:Uݳ,*W
z_KN_"&FkG$pn~A/eTw/ĥF(.1Xxr*IH7,Q 봺Of͊~2ux# Nq@:פqևX4u/rUA#%Jk{ݮOQaҵ(c[yVK8~Y}rZeTYW)zƁ)v/vs+̑V
U\F #zC8:٪B