Analysis of the Optimization Landscapes for Overcomplete Representation Learning

Total Page:16

File Type:pdf, Size:1020Kb

Analysis of the Optimization Landscapes for Overcomplete Representation Learning Analysis of the Optimization Landscapes for Overcomplete Representation Learning Qing Qu Center for Data Science New York University July 19, 2020 Data Increasingly Massive & High-Dimensional... hyperspectral imaging autonomous driving social network healthcare Learning Compact Representations Learning Compact Representations Summary of Main Results Convolutional/overcomplete dictionary learning can be provably solved with simple methods. Q. Qu, Y. Zhai, X. Li, Y. Zhang, Z. Zhu, Analysis of optimization landscapes for overcomplete learning, ICLR’20. Outline Overcomplete Dictionary Learning Convolutional Dictionary Learning Conclusion Learning Sparsely-Used Dictionaries Given Y , jointly learn compact dictionary A0 and sparse X0? Learning Compact Representations Denoising Image Restoration Super Resolution Image Half-toning - Image courtesy of Julien Mairal et al. Dictionary Learning - Symmetry ♦ Permutation symmetry: (2nn! signed permutations Π) ⊤ Y = A0X0 = (A0Π) Π X0 ⊤ ♦ Equivalent solution pairs: (A0, X0) ⇐⇒ A0Π, Π X0 . Dictionary Learning - Symmetry Leads to Nonconvexity ♦ Permutation symmetry: (2nn! signed permutations Π) ⊤ Y = A0X0 = (A0Π) Π X0 ⊤ ♦ Equivalent solution pairs: (A0, X0) ⇐⇒ A0Π, Π X0 . “flat” saddle point In the worst case, finding a local minimizer is NP-hard [Murty et al. 1987]... Why Do Nonconvex Problems Seem “Scary”? Naturally be formulated as nonconvex optimization problems: min φ( Y ; W ) W ∈M data model parameters “bad” local minima In the worst case, finding a local minimizer is NP-hard [Murty et al. 1987]... Why Do Nonconvex Problems Seem “Scary”? Naturally be formulated as nonconvex optimization problems: min φ( Y ; W ) W ∈M data model parameters “bad” local minima “flat” saddle point Why Do Nonconvex Problems Seem “Scary”? Naturally be formulated as nonconvex optimization problems: min φ( Y ; W ) W ∈M data model parameters “bad” local minima “flat” saddle point In the worst case, finding a local minimizer is NP-hard [Murty et al. 1987]... ♦ No “bad” local minima; ♦ No “flat” saddle point. Symmetry Creates Benign Nonconvex Geometry Symmetry Creates Benign Nonconvex Geometry ♦ No “bad” local minima; ♦ No “flat” saddle point. Symmetry Creates Benign Nonconvex Geometry Nonconvex learning problems can be solved efficiently to global solutions! ♦ sparse blind deconvolution [Li et al.’18, Qu et al.’19] ♦ overcomplete dictionary learning [this work] ♦ convolutional dictionary learning [this work] ♦ tensor decomposition [Ge et al.’16] A Fairly Broad Class of Nonconvex Problems ♦ phase retrieval [Sun et al’18] ♦ low rank matrix recovery [Ge et al.’16, Zhu et al.’18] ♦ phase synchronization [Boumal’17] ♦ shallow/linear neural network [Kawaguchi’17, Du et al.’19] A Fairly Broad Class of Nonconvex Problems ♦ sparse blind deconvolution ♦ phase retrieval [Sun et al’18] [Li et al.’18, Qu et al.’19] ♦ low rank matrix recovery ♦ overcomplete dictionary [Ge et al.’16, Zhu et al.’18] learning [this work] ♦ phase synchronization ♦ convolutional dictionary [Boumal’17] learning [this work] ♦ shallow/linear neural network ♦ tensor decomposition [Kawaguchi’17, Du et al.’19] [Ge et al.’16] Global Geometry for Overcomplete DL Given Y = A0 · X0, learn overcomplete A0 and sparse X0? Global Geometry for Overcomplete DL Given Y = A0 · X0, learn overcomplete A0 and sparse X0? Global Geometry for Overcomplete DL Given Y = A0 · X0, learn overcomplete A0 and sparse X0? Global Geometry for Overcomplete DL Given Y = A0 · X0, learn overcomplete A0 and sparse X0? Global Geometry for Overcomplete DL Find one column of A0 by ⊤ 4 min φDL(q) := − q Y q∈Sn−1 4 Theorem (Global Geometry, Qu et al.’19) Suppose (i) K = m/n constant (ii) A0 near orthogonal. Every critical point of φDL(q) is either 1. a saddle point exhibits negative curvature; 2. close to a target solution: one column of A0. Assumptions - Overcomplete DL Assumptions: near orthogonality of A0 1. Row orthogonal: unit norm tight frame (UNTF) r n A A⊤ = I, ∥a ∥ = 1. m 0 0 0i 2. Column (near) orthogonal: incoherence * + a a 0i 0j ≤ max , µ, i≠ j ∥a0i∥ ∥a0j∥ High-level Intuition: for any q⋆ = a0k h i E φ(q⊤Y ) ∝ φ( q⊤A ) X0 ⋆ | ⋆{z 0} correlation ζ(q⋆) Complete DL Overcomplete DL · ∥·∥ · − ∥·∥4 φ( ) = 1 φ( ) = 4 minimize sparsity on ζ(q) maximize spikiness on ζ(q) ρ(ζ) = #NNZ ρ(ζ) = ζ(1)/ζ(2) 3a. Li et al. Global geometry of multichannel sparse blind deconvolution on the sphere, NeurIPS’18. 3b. Zhang et al. Structured local optima in sparse blind deconvolution problem, IEEE TIT’19. 3c. Zhai et al., Complete dictionary learning via ℓ4-Norm maximization over the orthogonal group, JMLR’20. Problem Formulation - Overcomplete DL 1 Find one column of A0 by 4 ⊤ n−1 min φDL(q) := − q Y , s.t. q ∈ S . q 4 remove scaling spikiness promoting Complete DL Overcomplete DL · ∥·∥ · − ∥·∥4 φ( ) = 1 φ( ) = 4 minimize sparsity on ζ(q) maximize spikiness on ζ(q) ρ(ζ) = #NNZ ρ(ζ) = ζ(1)/ζ(2) 3a. Li et al. Global geometry of multichannel sparse blind deconvolution on the sphere, NeurIPS’18. 3b. Zhang et al. Structured local optima in sparse blind deconvolution problem, IEEE TIT’19. 3c. Zhai et al., Complete dictionary learning via ℓ4-Norm maximization over the orthogonal group, JMLR’20. Problem Formulation - Overcomplete DL 1 Find one column of A0 by 4 ⊤ n−1 min φDL(q) := − q Y , s.t. q ∈ S . q 4 remove scaling spikiness promoting High-level Intuition: for any q⋆ = a0k h i E φ(q⊤Y ) ∝ φ( q⊤A ) X0 ⋆ | ⋆{z 0} correlation ζ(q⋆) Problem Formulation - Overcomplete DL 1 Find one column of A0 by 4 ⊤ n−1 min φDL(q) := − q Y , s.t. q ∈ S . q 4 remove scaling spikiness promoting High-level Intuition: for any q⋆ = a0k h i E φ(q⊤Y ) ∝ φ( q⊤A ) X0 ⋆ | ⋆{z 0} correlation ζ(q⋆) Complete DL Overcomplete DL · ∥·∥ · − ∥·∥4 φ( ) = 1 φ( ) = 4 minimize sparsity on ζ(q) maximize spikiness on ζ(q) ρ(ζ) = #NNZ ρ(ζ) = ζ(1)/ζ(2) 3a. Li et al. Global geometry of multichannel sparse blind deconvolution on the sphere, NeurIPS’18. 3b. Zhang et al. Structured local optima in sparse blind deconvolution problem, IEEE TIT’19. 3c. Zhai et al., Complete dictionary learning via ℓ4-Norm maximization over the orthogonal group, JMLR’20. Problem Formulation - Overcomplete DL If q⋆ = a01, as the columns of A0 are incoherent: h i ζ(q ) = 1 a⊤ a ··· a⊤ a ⋆ | 01{z 02} | 01{z 02} |·|<µ |·|<µ Complete DL Overcomplete DL · ∥·∥ · − ∥·∥4 φ( ) = 1 φ( ) = 4 minimize sparsity on ζ(q) maximize spikiness on ζ(q) ρ(ζ) = #NNZ ρ(ζ) = ζ(1)/ζ(2) Global Geometry for Overcomplete DL Find one column of A0 by ⊤ 4 min φDL(q) := − q Y q∈Sn−1 4 Theorem (Global Geometry, Qu et al.’19) Suppose (i) K = m/n constant (ii) A0 near orthogonal. Every critical point of φDL(q) is either 1. a saddle point exhibits negative curvature; 2. close to a target solution: one column of A0. Global Geometry for Overcomplete DL Experiment - Overcomplete DL 2 practice m < n recover full A0 via repeated vs. theory m < Cn independent trials Choice of ℓk-Norm for Dictionary Learning k Average error with varying Maximizing ℓ -norm with sample complexity p/n and k. different k. Image credited to Yuexiang Zhai and Yifei Shen. - Zhai et al. Complete dictionary learning via ℓ4-norm maximization over the orthogonal group, JMLR’20. - Shen et al. Complete dictionary learning via ℓp-norm maximization, ArXiv preprint’20. Outline Overcomplete Dictionary Learning Convolutional Dictionary Learning Conclusion Convolutional Dictionary Learning Given multiple measurement yi of circulant convolution XK ~ ≤ ≤ yi = ak xki, (1 i p), k=1 { } { } can we jointly learn all ak k and sparse xki k,i? ♦ ∈ Rn Here, yi, ak, xki ♦ Sparse signal xki - Il Yong Chun and Jeffrey A. Fessler, Convolutional dictionary learning: Acceleration and convergence, TIP’18. - Cristina Garcia-Cardona and Brendt Wohlberg, Convolutional dictionary learning: A comparative review & new algorithms, TCI’18. Motivations Learning compact convolutional representations [Bristow et al.’13, Gu et al.’15, Garcia-Cardona et al.’18] Learning physical models from scientific data [Cheung et al.’18] Motivations Can be viewed as one-layer of convolutional neural network [Papyan et al.’17&18, Ye et al.’18]. - Papyan et al., Convolutional neural networks analyzed via convolutional sparse coding, JMLR’18. - Papyan et al., Theoretical foundations of deep learning via sparse representations: A multilayer sparse model and its connection to convolutional neural networks, SPM’18. Problem Formulation - Circulant Matrices For each ak ~ xki, equivalently ~ ⇐⇒ · ≤ ≤ ≤ ≤ ak xki Cak Cxki , 1 k K, 1 i p, where a circulant matrix Convolutional DL vs. Overcomplete DL For each sample (1 ≤ i ≤ p), XK ~ yi = ak xki, k=1 equivalently, Cx1i h i . Cy = Ca ··· Ca · . , i | 1 {z K } . overcomplete A0 Cx | {zKi } sparse Xi Convolutional DL vs. Overcomplete DL Given Y = A0 · X0, learn overcomplete A0 and sparse X0? - Huang et al., Convolutional dictionary learning through tensor factorization, NeurIPS Workshop’15. From Overcomplete DL to Convolutional DL? Find one shift of ai (1 ≤ i ≤ K) via 4 min − q⊤Y , s.t. q ∈ Sn−1 ? q 4 From Overcomplete DL to Convolutional DL? Find one shift of ai (1 ≤ i ≤ K) via 4 min − q⊤Y , s.t. q ∈ Sn−1 ? q 4 For generic ai and A0, it does NOT work! ♦ Generic ai enforces structures on A0; ♦ For convolutional DL, in general A0 is NOT near orthogonal. Problem Formulation - Convolutional DL 4 ⊤ n−1 min φCDL(q) = − q P Y , s.t. q ∈ S , q 4 ♦ Preconditioning matrix: −1/2 −1/2 −1 ⊤ ≈ ⊤ P = (θnK) YY A0A0 Problem Formulation - Convolutional DL 4 ⊤ n−1 min φCDL(q) ≈ − q AX0 , s.t. q ∈ S , q 4 ♦ Row orthogonalization of A0: −1/2 ≈ ⊤ PY A0A0 A0 X0 = AX0. | {z } A ♦ A is tight frame (but not necessarily unit norm).
Recommended publications
  • Landscape Analysis of Overcomplete Tensor and Neural Collapse
    Landscape Analysis of Overcomplete Tensor and Neural Collapse Qing Qu Dept. of EECS, University of Michigan May 17, 2021 Outline of this Talk • Introduction • Overcomplete Tensor Decomposition (Representation Learning) • Neural Collapse in Deep Network Training Outline of this Talk • Introduction • Overcomplete Tensor Decomposition (Representation Learning) • Neural Collapse in Deep Network Training Nonconvex Problems in Representation Learning 5/18/21 4 General Nonconvex Problems 5/18/21 5 General Nonconvex Problems 5/18/21 6 General Nonconvex Problems 5/18/21 7 Optimizing Nonconvex Problems Globally 5/18/21 8 Nonconvex Problems with Benign Landscape • Generalized Phase Retrieval [Sun’18] • Low-rank Matrix Recovery [Ma’16, Jin’17, Chi’19] • (Convolutional) Sparse Dictionary Learning [Sun’16, Qu’20] • (Orthogonal) Tensor Decomposition [Ge’15] • Sparse Blind Deconvolution [Zhang’17, Li’18, Kuo’19] • Deep Linear Network [Kawaguchi’16] • ... 5/18/21 9 Outline of this Talk • Introduction • Overcomplete Tensor Decomposition (Representation Learning) • Neural Collapse in Deep Network Training Landscape Analysis of Overcomplete Learning Q. Qu, Y. Zhai, X. Li, Y. Zhang, Z. Zhu, Analysis of optimization landscapes for overcomplete learning, ICLR’20, (oral, top 1.9%) • Provide the global landscape for overcomplete representation learning problems. • Explains why they can be efficiently optimized to global optimality Overcomplete Tensor Decomposition We consider decomposing a 4-th order tensor of rank m in the following form Core problem for several unsupervised representation learning problems (ICA and mixture of Gaussian [Anandkumar’12], dictionary learning [Barak’14,Qu’20]), and even training neural networks [Ge’17]. Overcomplete Tensor Decomposition A natural (nonconvex) objective to find one component Overcomplete Tensor Decomposition Overcomplete Tensor Decomposition • For overcomplete case, most of existing landscape analysis results [Ge’17] are local, or are based on Sum-of-Squares relaxations [Barak’15, Ma’16] which is computationally expensive.
    [Show full text]
  • Kūnqǔ in Practice: a Case Study
    KŪNQǓ IN PRACTICE: A CASE STUDY A DISSERTATION SUBMITTED TO THE GRADUATE DIVISION OF THE UNIVERSITY OF HAWAI‘I AT MĀNOA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY IN THEATRE OCTOBER 2019 By Ju-Hua Wei Dissertation Committee: Elizabeth A. Wichmann-Walczak, Chairperson Lurana Donnels O’Malley Kirstin A. Pauka Cathryn H. Clayton Shana J. Brown Keywords: kunqu, kunju, opera, performance, text, music, creation, practice, Wei Liangfu © 2019, Ju-Hua Wei ii ACKNOWLEDGEMENTS I wish to express my gratitude to the individuals who helped me in completion of my dissertation and on my journey of exploring the world of theatre and music: Shén Fúqìng 沈福庆 (1933-2013), for being a thoughtful teacher and a father figure. He taught me the spirit of jīngjù and demonstrated the ultimate fine art of jīngjù music and singing. He was an inspiration to all of us who learned from him. And to his spouse, Zhāng Qìnglán 张庆兰, for her motherly love during my jīngjù research in Nánjīng 南京. Sūn Jiàn’ān 孙建安, for being a great mentor to me, bringing me along on all occasions, introducing me to the production team which initiated the project for my dissertation, attending the kūnqǔ performances in which he was involved, meeting his kūnqǔ expert friends, listening to his music lessons, and more; anything which he thought might benefit my understanding of all aspects of kūnqǔ. I am grateful for all his support and his profound knowledge of kūnqǔ music composition. Wichmann-Walczak, Elizabeth, for her years of endeavor producing jīngjù productions in the US.
    [Show full text]
  • Is Shuma the Chinese Analog of Soma/Haoma? a Study of Early Contacts Between Indo-Iranians and Chinese
    SINO-PLATONIC PAPERS Number 216 October, 2011 Is Shuma the Chinese Analog of Soma/Haoma? A Study of Early Contacts between Indo-Iranians and Chinese by ZHANG He Victor H. Mair, Editor Sino-Platonic Papers Department of East Asian Languages and Civilizations University of Pennsylvania Philadelphia, PA 19104-6305 USA [email protected] www.sino-platonic.org SINO-PLATONIC PAPERS FOUNDED 1986 Editor-in-Chief VICTOR H. MAIR Associate Editors PAULA ROBERTS MARK SWOFFORD ISSN 2157-9679 (print) 2157-9687 (online) SINO-PLATONIC PAPERS is an occasional series dedicated to making available to specialists and the interested public the results of research that, because of its unconventional or controversial nature, might otherwise go unpublished. The editor-in-chief actively encourages younger, not yet well established, scholars and independent authors to submit manuscripts for consideration. Contributions in any of the major scholarly languages of the world, including romanized modern standard Mandarin (MSM) and Japanese, are acceptable. In special circumstances, papers written in one of the Sinitic topolects (fangyan) may be considered for publication. Although the chief focus of Sino-Platonic Papers is on the intercultural relations of China with other peoples, challenging and creative studies on a wide variety of philological subjects will be entertained. This series is not the place for safe, sober, and stodgy presentations. Sino- Platonic Papers prefers lively work that, while taking reasonable risks to advance the field, capitalizes on brilliant new insights into the development of civilization. Submissions are regularly sent out to be refereed, and extensive editorial suggestions for revision may be offered. Sino-Platonic Papers emphasizes substance over form.
    [Show full text]
  • The Linguistic Categorization of Deictic Direction in Chinese – with Reference to Japanese – Christine Lamarre
    The linguistic categorization of deictic direction in Chinese – With reference to Japanese – Christine Lamarre To cite this version: Christine Lamarre. The linguistic categorization of deictic direction in Chinese – With reference to Japanese –. Dan XU. Space in Languages of China, Springer, pp.69-97, 2008, 978-1-4020-8320-4. hal-01382316 HAL Id: hal-01382316 https://hal-inalco.archives-ouvertes.fr/hal-01382316 Submitted on 16 Oct 2016 HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est archive for the deposit and dissemination of sci- destinée au dépôt et à la diffusion de documents entific research documents, whether they are pub- scientifiques de niveau recherche, publiés ou non, lished or not. The documents may come from émanant des établissements d’enseignement et de teaching and research institutions in France or recherche français ou étrangers, des laboratoires abroad, or from public or private research centers. publics ou privés. Lamarre, Christine. 2008. The linguistic categorization of deictic direction in Chinese — With reference to Japanese. In Dan XU (ed.) Space in languages of China: Cross-linguistic, synchronic and diachronic perspectives. Berlin/Heidelberg/New York: Springer, pp.69-97. THE LINGUISTIC CATEGORIZATION OF DEICTIC DIRECTION IN CHINESE —— WITH REFERENCE TO JAPANESE —— Christine Lamarre, University of Tokyo Abstract This paper discusses the linguistic categorization of deictic direction in Mandarin Chinese, with reference to Japanese. It focuses on the following question: to what extent should the prevalent bimorphemic (nondeictic + deictic) structure of Chinese directionals be linked to its typological features as a satellite-framed language? We know from other satellite-framed languages such as English, Hungarian, and Russian that this feature is not necessarily directly connected to satellite-framed patterns.
    [Show full text]
  • Ideophones in Middle Chinese
    KU LEUVEN FACULTY OF ARTS BLIJDE INKOMSTSTRAAT 21 BOX 3301 3000 LEUVEN, BELGIË ! Ideophones in Middle Chinese: A Typological Study of a Tang Dynasty Poetic Corpus Thomas'Van'Hoey' ' Presented(in(fulfilment(of(the(requirements(for(the(degree(of(( Master(of(Arts(in(Linguistics( ( Supervisor:(prof.(dr.(Jean=Christophe(Verstraete((promotor)( ( ( Academic(year(2014=2015 149(431(characters Abstract (English) Ideophones in Middle Chinese: A Typological Study of a Tang Dynasty Poetic Corpus Thomas Van Hoey This M.A. thesis investigates ideophones in Tang dynasty (618-907 AD) Middle Chinese (Sinitic, Sino- Tibetan) from a typological perspective. Ideophones are defined as a set of words that are phonologically and morphologically marked and depict some form of sensory image (Dingemanse 2011b). Middle Chinese has a large body of ideophones, whose domains range from the depiction of sound, movement, visual and other external senses to the depiction of internal senses (cf. Dingemanse 2012a). There is some work on modern variants of Sinitic languages (cf. Mok 2001; Bodomo 2006; de Sousa 2008; de Sousa 2011; Meng 2012; Wu 2014), but so far, there is no encompassing study of ideophones of a stage in the historical development of Sinitic languages. The purpose of this study is to develop a descriptive model for ideophones in Middle Chinese, which is compatible with what we know about them cross-linguistically. The main research question of this study is “what are the phonological, morphological, semantic and syntactic features of ideophones in Middle Chinese?” This question is studied in terms of three parameters, viz. the parameters of form, of meaning and of use.
    [Show full text]
  • <I>Comment Vous Appelez-Vous</I>?: Why the French Change Their Names
    Comment vous appelez-vous?: Why the French Change Their Names James E. Jacob California State University, Chico and Pierre L. Horn Wright State University We examine the history, processes, and motivations which explain in part why and how the French change their names. Studying a small but representative sample of official name changes mandatorily published in the JournaL OfficieL de La Republique Franfaise at certain moments of post-World War II history (1946, 1963, and 1992-1995), we identify the four most important reasons for changing names: because they are obscene or. pejorative; ridiculous; perceived as too foreign, especially too Arab or too Jewish; or to add the patent of nobility . While we express surprise that names of these kinds still exist in contemporary France, we also recognize that the decision to request a name change at this late date is the last refuge for families who have long suffered indignities and humiliations because of their names. The name I bear is not only a "gift" from my father or my mother; it is an integral part of myself; it has defined me ... since my childhood. Francois eros [T]he name is a family possession; it is the cement of the family. Marianne Mulon Names 46.1 (March 1998):3-28 ISSN:0027-7738 © 1998by The American Name Society 3 4 Names 46.1 (March 1998) With the Ordonnance de Villers-Cotterets in 1539, which, according to Albert Dauzat (1949, 40), codified a long-established custom of assigning and preserving family names, the French monarchy sought to further integrate the diversity of its realm.
    [Show full text]
  • Chinese Language and Characters
    Chinese Language and Characters Pronunciation of Chinese Words Consonants Pinyin WadeGiles Pronunciation Example: Pinyin(WadeGiles) Aspirated: p p’ pin Pao (P’ao) t t’ tip Tao (T’ao) k k’ kilt Kuan (K’uan) ch ch’ ch in, ch urch Chi (Ch’i) q ch’ ch eek Qi (Ch’i) c ts’ bi ts Cang (Ts’ang) Un- b p bin Bao (Pao) aspirated: d t dip Dao (Tao) g k gilt Guan (Kuan) r j wr en Ren (Jen) sh sh sh ore Shang (Shang) si szu Si (Szu) x hs or sh sh oe Xu (Hsu) z ts or tz bi ds Zang (Tsang) zh ch gin Zhong (Chong) zh j jeep Zhong (Jong) zi tzu Zi (Tzu) Vowels - a a father usually Italian e e ei ght values eh eh broth er yi i mach ine, p in Yi (I) i ih sh ir t Zhi (Chih) o soap u goo se ü über Dipthongs ai light ao lou d ei wei ght ia Will ia m ieh Kor ea ou gr ou p ua swa n ueh do er ui sway Hui (Hui) uo Whoah ! Combinations ian ien Tian (Tien) ui wei Wei gh Shui (Shwei) an and ang bun and b ung en and eng wood en and am ong in and ing sin and s ing ong un and ung u as in l oo k Tong (T’ung) you yu Watts, Alan; Tao The Watercourse Way, Pelican Books, 1976 http://acc6.its.brooklyn.cuny.edu/~phalsall/texts/chinlng1.html Tones 1 2 3 4 ā á ă à ē é ĕ È è Ī ī í ĭ ì ō ó ŏ ò ū ú ŭ ù Pinyin (Wade Giles) Meaning Ai Bā (Pa) Eight, see Numbers Bái (Pai) White, plain, unadorned Băi (Pai) One hundred, see Numbers Bāo Envelop Bāo (Pao) Uterus, afterbirth Bēi Sad, Sorrow, melancholy Bĕn Root, origin (Biao and Ben) see Biao Bi Bi (bei) Bian Bi āo Tip, dart, javelin, (Biao and Ben) see Ben Bin Bin Bing Bu Bu Can Cang Cáng (Ts’ang) Hidden, concealed (see Zang) Cháng Intestine Ch ōng (Ch’ung) Surging Ch ōng (Ch’ung) Rushing Chóu Worry Cóng Follow, accord with Dăn (Tan) Niche or shrine Dăn (Tan) Gall Bladder Dān (Tan) Red Cinnabar Dào (Tao) The Way Dì (Ti) The Earth, i.e.
    [Show full text]
  • Origin Narratives: Reading and Reverence in Late-Ming China
    Origin Narratives: Reading and Reverence in Late-Ming China Noga Ganany Submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy in the Graduate School of Arts and Sciences COLUMBIA UNIVERSITY 2018 © 2018 Noga Ganany All rights reserved ABSTRACT Origin Narratives: Reading and Reverence in Late Ming China Noga Ganany In this dissertation, I examine a genre of commercially-published, illustrated hagiographical books. Recounting the life stories of some of China’s most beloved cultural icons, from Confucius to Guanyin, I term these hagiographical books “origin narratives” (chushen zhuan 出身傳). Weaving a plethora of legends and ritual traditions into the new “vernacular” xiaoshuo format, origin narratives offered comprehensive portrayals of gods, sages, and immortals in narrative form, and were marketed to a general, lay readership. Their narratives were often accompanied by additional materials (or “paratexts”), such as worship manuals, advertisements for temples, and messages from the gods themselves, that reveal the intimate connection of these books to contemporaneous cultic reverence of their protagonists. The content and composition of origin narratives reflect the extensive range of possibilities of late-Ming xiaoshuo narrative writing, challenging our understanding of reading. I argue that origin narratives functioned as entertaining and informative encyclopedic sourcebooks that consolidated all knowledge about their protagonists, from their hagiographies to their ritual traditions. Origin narratives also alert us to the hagiographical substrate in late-imperial literature and religious practice, wherein widely-revered figures played multiple roles in the culture. The reverence of these cultural icons was constructed through the relationship between what I call the Three Ps: their personas (and life stories), the practices surrounding their lore, and the places associated with them (or “sacred geographies”).
    [Show full text]
  • Gěi ’Give’ in Beijing and Beyond Ekaterina Chirkova
    Gěi ’give’ in Beijing and beyond Ekaterina Chirkova To cite this version: Ekaterina Chirkova. Gěi ’give’ in Beijing and beyond. Cahiers de linguistique - Asie Orientale, CRLAO, 2008, 37 (1), pp.3-42. hal-00336148 HAL Id: hal-00336148 https://hal.archives-ouvertes.fr/hal-00336148 Submitted on 2 Nov 2008 HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est archive for the deposit and dissemination of sci- destinée au dépôt et à la diffusion de documents entific research documents, whether they are pub- scientifiques de niveau recherche, publiés ou non, lished or not. The documents may come from émanant des établissements d’enseignement et de teaching and research institutions in France or recherche français ou étrangers, des laboratoires abroad, or from public or private research centers. publics ou privés. Gěi ‘give’ in Beijing and beyond1 Katia Chirkova (CRLAO, CNRS) This article focuses on the various uses of gěi ‘give’, as attested in a corpus of spoken Beijing Mandarin collected by the author. These uses are compared to those in earlier attestations of Beijing Mandarin and to those in Greater Beijing Mandarin and in Jì-Lǔ Mandarin dialects. The uses of gěi in the corpus are demonstrated to be consistent with the latter pattern, where the primary function of gěi is that of indirect object marking and where, unlike Standard Mandarin, gěi is not additionally used as an agent marker or a direct object marker. Exceptions to this pattern in the corpus are explained as a recent development arisen through reanalysis. Key words : gěi, direct object marker, indirect object marker, agent marker, Beijing Mandarin, Northern Mandarin, typology.
    [Show full text]
  • Linguistic Composition and Characteristics of Chinese Given Names DOI: 10.34158/ONOMA.51/2016/8
    Onoma 51 Journal of the International Council of Onomastic Sciences ISSN: 0078-463X; e-ISSN: 1783-1644 Journal homepage: https://onomajournal.org/ Linguistic composition and characteristics of Chinese given names DOI: 10.34158/ONOMA.51/2016/8 Irena Kałużyńska Sinology Department Faculty of Oriental Studies University of Warsaw e-mail: [email protected] To cite this article: Kałużyńska, Irena. 2016. Linguistic composition and characteristics of Chinese given names. Onoma 51, 161–186. DOI: 10.34158/ONOMA.51/2016/8 To link to this article: https://doi.org/10.34158/ONOMA.51/2016/8 © Onoma and the author. Linguistic composition and characteristics of Chinese given names Abstract: The aim of this paper is to discuss various linguistic and cultural aspect of personal naming in China. In Chinese civilization, personal names, especially given names, were considered crucial for a person’s fate and achievements. The more important the position of a person, the more various categories of names the person received. Chinese naming practices do not restrict the inventory of possible given names, i.e. given names are formed individually, mainly as a result of a process of onymisation, and given names are predominantly semantically transparent. Therefore, given names seem to be well suited for a study of stereotyped cultural expectations present in Chinese society. The paper deals with numerous subdivisions within the superordinate category of personal name, as the subclasses of surname and given name. It presents various subcategories of names that have been used throughout Chinese history, their linguistic characteristics, their period of origin, and their cultural or social functions.
    [Show full text]
  • The Impact of Emergency Surgery Versus Self
    The Impact of Emergency Surgery versus Self- expanding Metallic Stents on Anal Function and Quality of Life in Patients with Complete Obstructing Left-sided Colon Cancer GanBin Li ( [email protected] ) Beijing Chaoyang Hospital https://orcid.org/0000-0002-2576-1100 Tao Yu Beijing Chaoyang Hospital Hao-Yu Zhang Beijing Chaoyang Hospital Zhen-Jun Wang Beijing Chaoyang Hospital Guang-Hui Wei Beijing Chao-Yang Hospital: Beijing Chaoyang Hospital Hao Qu Beijing Chaoyang Hospital Zhai-Wei Zhai Beijing Chaoyang Hospital Jia Gang Han Beijing Chaoyang Hospital Research article Keywords: Complete obstructing left-sided colon cancer, Emergency surgery, Self-expanding metallic stents, Anal function, Quality of life Posted Date: February 11th, 2021 DOI: https://doi.org/10.21203/rs.3.rs-208724/v1 License: This work is licensed under a Creative Commons Attribution 4.0 International License. Read Full License Page 1/16 Abstract Background: Emergence surgery (ES) and self-expanding metallic stents (SEMS) are traditional approaches for complete obstructing left-sided colon cancer. A strategy of “stents-chemotherapy-surgery” was applied in our center recently. Studies assessing the anal function and quality of life of patients with complete obstructing left-sided colon cancer are still lacking. Methods: Patients with complete obstructing left-sided colon cancer were included, and three treatment strategies were used, including ES, SEMS, and SEMS followed by neoadjuvant chemotherapy (NAC) for patients with complete obstructing left-sided colon cancer. The Wexner, Vaizey, and low anterior resection syndrome (LARS) scores were used to assess anal function and the EORTC QLQ C30 score was used to assess quality of life.
    [Show full text]
  • Representing Talented Women in Eighteenth-Century Chinese Painting: Thirteen Female Disciples Seeking Instruction at the Lake Pavilion
    REPRESENTING TALENTED WOMEN IN EIGHTEENTH-CENTURY CHINESE PAINTING: THIRTEEN FEMALE DISCIPLES SEEKING INSTRUCTION AT THE LAKE PAVILION By Copyright 2016 Janet C. Chen Submitted to the graduate degree program in Art History and the Graduate Faculty of the University of Kansas in partial fulfillment of the requirements for the degree of Doctor of Philosophy. ________________________________ Chairperson Marsha Haufler ________________________________ Amy McNair ________________________________ Sherry Fowler ________________________________ Jungsil Jenny Lee ________________________________ Keith McMahon Date Defended: May 13, 2016 The Dissertation Committee for Janet C. Chen certifies that this is the approved version of the following dissertation: REPRESENTING TALENTED WOMEN IN EIGHTEENTH-CENTURY CHINESE PAINTING: THIRTEEN FEMALE DISCIPLES SEEKING INSTRUCTION AT THE LAKE PAVILION ________________________________ Chairperson Marsha Haufler Date approved: May 13, 2016 ii Abstract As the first comprehensive art-historical study of the Qing poet Yuan Mei (1716–97) and the female intellectuals in his circle, this dissertation examines the depictions of these women in an eighteenth-century handscroll, Thirteen Female Disciples Seeking Instructions at the Lake Pavilion, related paintings, and the accompanying inscriptions. Created when an increasing number of women turned to the scholarly arts, in particular painting and poetry, these paintings documented the more receptive attitude of literati toward talented women and their support in the social and artistic lives of female intellectuals. These pictures show the women cultivating themselves through literati activities and poetic meditation in nature or gardens, common tropes in portraits of male scholars. The predominantly male patrons, painters, and colophon authors all took part in the formation of the women’s public identities as poets and artists; the first two determined the visual representations, and the third, through writings, confirmed and elaborated on the designated identities.
    [Show full text]