東華大學圖書館 |

語系: 繁體中文

說明(常見問題)

回圖書館首頁

手機版館藏查詢

登入

回首頁 到查詢結果 [ null ]

切換: 標籤 | MARC模式 | ISBD

Advancing Sequence Modeling With Dee...

Smith, James Thomas Howard,

FindBook

Google Book

Amazon

博客來

Advancing Sequence Modeling With Deep State Space Methods /

紀錄類型:	書目-電子資源 : Monograph/item
正題名/作者:	Advancing Sequence Modeling With Deep State Space Methods // James Thomas Howard Smith.
作者:	Smith, James Thomas Howard,
面頁冊數:	1 electronic resource (216 pages)
附註:	Source: Dissertations Abstracts International, Volume: 86-07, Section: B.
Contained By:	Dissertations Abstracts International86-07B.
標題:	Deep learning. -
電子資源:	https://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=31821810
ISBN:	9798302377395

Advancing Sequence Modeling With Deep State Space Methods /
Smith, James Thomas Howard,

Advancing Sequence Modeling With Deep State Space Methods /James Thomas Howard Smith. - 1 electronic resource (216 pages)

Source: Dissertations Abstracts International, Volume: 86-07, Section: B.

This thesis advances sequence modeling through novel contributions bridging state space models and deep learning. Sequence modeling techniques identify patterns and relationships in ordered data, enabling diverse applications including predicting time series, processing speech and audio signals, generating text, modeling biological sequences, and analyzing neural recordings. While sequence modeling has been transformed by deep learning, current architectures face fundamental tradeoffs between computational efficiency, modeling capability, and interpretability. We address these challenges through three complementary contributions.We first introduce the S5 layer, which redesigns deep state space methods to enable parallel scan operations. We establish mathematical connections with prior methods that explain its strong empirical performance. We then extend these insights to spatiotemporal modeling with ConvS5, which integrates spatial structure into the state space framework. Finally, we develop new analytical tools through the Jacobian Switching Linear Dynamical System (JSLDS) procedure, providing a systematic way to decompose complex nonlinear dynamics into interpretable linear components.Through extensive empirical validation across a variety of domains, we show how these methods significantly advance the state-of-the-art in their respective areas. Together, these contributions demonstrate the broad potential of combining classical dynamical systems approaches with modern deep learning techniques to create more capable, efficient, and interpretable sequence models.

English

ISBN: 9798302377395Subjects--Topical Terms:

3554982
Deep learning.

Advancing Sequence Modeling With Deep State Space Methods /
LDR:02903nmm a22003853i 4500 001 2432625
005 20250731101859.5
006 m o d
007 cr|nu||||||||
008 261011s2024 miu||||||m |||||||eng d
020 $a 9798302377395
035 $a (MiAaPQD)AAI31821810
035 $a (MiAaPQD)Stanfordgz824mn4488
035 $a AAI31821810
040 $a MiAaPQD $b eng $c MiAaPQD $e rda
100 1 $a Smith, James Thomas Howard, $e author. $3 3821920
245 1 0 $a Advancing Sequence Modeling With Deep State Space Methods / $c James Thomas Howard Smith.
264 1 $a Ann Arbor : $b ProQuest Dissertations & Theses, $c 2024
300 $a 1 electronic resource (216 pages)
336 $a text $b txt $2 rdacontent
337 $a computer $b c $2 rdamedia
338 $a online resource $b cr $2 rdacarrier
500 $a Source: Dissertations Abstracts International, Volume: 86-07, Section: B.
500 $a Advisors: Linderman, Scott Committee members: Palacios, Julia; Ré, Chris; Bent, Stacey F.
502 $b Ph.D. $c Stanford University $d 2024.
520 $a This thesis advances sequence modeling through novel contributions bridging state space models and deep learning. Sequence modeling techniques identify patterns and relationships in ordered data, enabling diverse applications including predicting time series, processing speech and audio signals, generating text, modeling biological sequences, and analyzing neural recordings. While sequence modeling has been transformed by deep learning, current architectures face fundamental tradeoffs between computational efficiency, modeling capability, and interpretability. We address these challenges through three complementary contributions.We first introduce the S5 layer, which redesigns deep state space methods to enable parallel scan operations. We establish mathematical connections with prior methods that explain its strong empirical performance. We then extend these insights to spatiotemporal modeling with ConvS5, which integrates spatial structure into the state space framework. Finally, we develop new analytical tools through the Jacobian Switching Linear Dynamical System (JSLDS) procedure, providing a systematic way to decompose complex nonlinear dynamics into interpretable linear components.Through extensive empirical validation across a variety of domains, we show how these methods significantly advance the state-of-the-art in their respective areas. Together, these contributions demonstrate the broad potential of combining classical dynamical systems approaches with modern deep learning techniques to create more capable, efficient, and interpretable sequence models.
546 $a English
590 $a School code: 0212
650 4 $a Deep learning. $3 3554982
650 4 $a Dynamical systems. $3 3219613
650 4 $a Speech. $3 530580
650 4 $a Benchmarks. $3 3548256
650 4 $a Mathematics. $3 515831
650 4 $a Engineering. $3 586835
690 $a 0800
690 $a 0405
690 $a 0537
710 2 $a Stanford University. $e degree granting institution. $3 3765820
720 1 $a Linderman, Scott $e degree supervisor.
773 0 $t Dissertations Abstracts International $g 86-07B.
790 $a 0212
791 $a Ph.D.
792 $a 2024
856 4 0 $u https://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=31821810