1,721,020 research outputs found
Going Beyond Counting First Authors in Author Co-citation Analysis
Get PDFThe present study examines one of the fundamental aspects of author co-citation analysis (ACA) - the way co-citation
counts are defined. Co-citation counting provides the data on which all subsequent statistical analyses and mappings
are based, and we compare ACA results based on two different types of co-citation counting - the traditional type that
only counts the first one among a cited work's authors on the one hand and a non-traditional type that takes into
account the first 5 authors of a cited work on the other hand. Results indicate that the picture produced through this non-traditional author co-citation counting contains more coherent author groups and is therefore considerably clearer. However, this picture represents fewer specialties in the research field being studied than that produced through the traditional first-author co-citation counting when the same number of top-ranked authors is selected and analyzed. Reasons for these effects are discussed
Variations on the Author
Get PDF“Variations on the Author” discusses two of Eduardo Coutinho’s recent films (Um Dia na Vida, from 2010, and Últimas Conversas, posthumously released in 2015) and their contribution to the general question of documentary authorship. The director’s filmography is characterized by a consistent yet self-effacing form of authorial self-inscription: Coutinho often features as an interviewer that rather than express opinions propels discourses; an interviewer that is good at listening. This mode of self-inscription characterizes him as an author who is not expressive but who is nonetheless markedly present on the screen. In Um Dia na Vida, however, Coutinho is completely absent form the image, while Últimas Conversas, on the contrary, includes a confessional prologue that moves the director from the margins to the center of his films. This article examines the ways in which these works stand out in the filmography of a director who offers new insights into the notion of cinematic authorship
Appropriate Similarity Measures for Author Cocitation Analysis
Get PDFWe provide a number of new insights into the methodological discussion about author cocitation analysis. We first argue that the use of the Pearson correlation for measuring the similarity between authors’ cocitation profiles is not very satisfactory. We then discuss what kind of similarity measures may be used as an alternative to the Pearson correlation. We consider three similarity measures in particular. One is the well-known cosine. The other two similarity measures have not been used before in the bibliometric literature. Finally, we show by means of an example that our findings have a high practical relevance.information science;Pearson correlation;cosine;similarity measure;author cocitation analysis
On the diagnosable, efficient, scalable and effective learning
Get PDFDeep learning is a cornerstone in the quest for Artificial General Intelligence (AGI), offering a transformative approach to machines' comprehension of and interaction with the world. At its core, deep learning mimics the complexity and adaptability of human intelligence by processing and learning from vast amounts of data through neural networks. This approach enables machines to perform a wide range of tasks, such as computer vision, natural language processing, and speech recognition, with unprecedented flexibility and learning capacity.
A mature deep learning system comprises a reliable evaluation method, a strong model, and an effective training approach. In this dissertation, I focus on developing algorithms from these three perspectives. For the evaluation, (i) we propose PatchAttack, a black-box adversarial attack algorithm, to diagnose whether a deep network is robust against localized changes that do not confuse humans. For the model, (ii) we introduce the Lite Vision Transformer (LVT), a mobile transformer designed to tackle the problem of insufficient representation capability with limited model parameters. (iii) We propose a model series called MOAT to explore the design principles of scalable vision transformers that benefit both upstream and downstream tasks. For the training, (iv) we design a knowledge distillation algorithm to train neural networks across multiple generations using the same architecture, aiming to explore the benefits of self-supervision. (v) We further propose the Snapshot Distillation algorithm to condense the knowledge distillation process across multiple model generations into a single generation. (vi) Finally, we study the training algorithm for the vision-language model and propose the Information Gain (IG) Captioner, a multimodal GPT model, to explore the benefits of the information gain training method on zero-shot tasks
Robust 3D Recognition via Analysis-by-Synthesis
No full textImages are 2D observations of the 3D world we live in. Recognition of objects from those images is one of the most important tasks for computer vision. Currently, the majority of computer vision approaches recognize 2D images without modeling the imaging process from 3D space. These approaches learn from the distribution of the train 2D images and are normally sensitive to out-of-distribution (OOD) situations. Although we have developed an alternative approach that significantly improves the robustness of 2D vision approaches, we believe it is still limited. Motivated by cognitive studies of human vision, we propose an object recognition pipeline, which recognizes objects from 2D images by predicting the 3D configuration in the 3D space while modeling the imaging process. Our proposed approach learns approximate category-level object representation by a combination of 3D object geometry and discriminative neural features. One crucial contribution of our proposed approach is the mechanism that bridges the 2D image and 3D world, which is namely Render-and-Compare. Specifically, Render-and-Compare conducts Analysis-by-Synthesis via differentiating the rendering process that provides cues for retrieval of 3D configurations of the objects in an iterative optimization process. On the other hand, our approach builds a flexible object centric representation. Using variant object geometry representations, optimization strategies, rendering strategies, and neural representations, our approach can be applied to lots of vision recognition tasks, including object pose estimation, shape estimation, detection, amodal segmentation, part detection, classification, and few-shot pose learning. During my path to pursuing PhD, I, with collaboration of others, have developed this approach and extensively studied all components in the pipeline for boarding its application with better accuracy and robustness. The extensive experiments conducted on various datasets demonstrate our approach ability for real-world usage
Dispelling the Myths Behind First-author Citation Counts
Get PDFWe conducted a full-scale evaluative citation analysis study of scholars in the XML research field to explore just how different from each other author rankings resulting from different citation counting methods actually are, and to demonstrate the capability of emerging data and tools on the Web in supporting more realistic citation counting methods. Our results contest some common arguments for the continued
use of first-author citation counts in the evaluation of scholars, such as high correlations between author rankings by first-author citation counts and other citation
counting methods, and high costs of using more realistic citation counting methods that are not well-supported by the ISI databases. It is argued that increasingly available digital full text research papers make it possible for citation analysis studies to go beyond what the ISI databases have directly supported and to employ more
sophisticated methods
On the diagnosable, efficient, scalable and effective learning
Get PDFDeep learning is a cornerstone in the quest for Artificial General Intelligence (AGI), offering a transformative approach to machines' comprehension of and interaction with the world. At its core, deep learning mimics the complexity and adaptability of human intelligence by processing and learning from vast amounts of data through neural networks. This approach enables machines to perform a wide range of tasks, such as computer vision, natural language processing, and speech recognition, with unprecedented flexibility and learning capacity.
A mature deep learning system comprises a reliable evaluation method, a strong model, and an effective training approach. In this dissertation, I focus on developing algorithms from these three perspectives. For the evaluation, (i) we propose PatchAttack, a black-box adversarial attack algorithm, to diagnose whether a deep network is robust against localized changes that do not confuse humans. For the model, (ii) we introduce the Lite Vision Transformer (LVT), a mobile transformer designed to tackle the problem of insufficient representation capability with limited model parameters. (iii) We propose a model series called MOAT to explore the design principles of scalable vision transformers that benefit both upstream and downstream tasks. For the training, (iv) we design a knowledge distillation algorithm to train neural networks across multiple generations using the same architecture, aiming to explore the benefits of self-supervision. (v) We further propose the Snapshot Distillation algorithm to condense the knowledge distillation process across multiple model generations into a single generation. (vi) Finally, we study the training algorithm for the vision-language model and propose the Information Gain (IG) Captioner, a multimodal GPT model, to explore the benefits of the information gain training method on zero-shot tasks
Robust 3D Recognition via Analysis-by-Synthesis
No full textImages are 2D observations of the 3D world we live in. Recognition of objects from those images is one of the most important tasks for computer vision. Currently, the majority of computer vision approaches recognize 2D images without modeling the imaging process from 3D space. These approaches learn from the distribution of the train 2D images and are normally sensitive to out-of-distribution (OOD) situations. Although we have developed an alternative approach that significantly improves the robustness of 2D vision approaches, we believe it is still limited. Motivated by cognitive studies of human vision, we propose an object recognition pipeline, which recognizes objects from 2D images by predicting the 3D configuration in the 3D space while modeling the imaging process. Our proposed approach learns approximate category-level object representation by a combination of 3D object geometry and discriminative neural features. One crucial contribution of our proposed approach is the mechanism that bridges the 2D image and 3D world, which is namely Render-and-Compare. Specifically, Render-and-Compare conducts Analysis-by-Synthesis via differentiating the rendering process that provides cues for retrieval of 3D configurations of the objects in an iterative optimization process. On the other hand, our approach builds a flexible object centric representation. Using variant object geometry representations, optimization strategies, rendering strategies, and neural representations, our approach can be applied to lots of vision recognition tasks, including object pose estimation, shape estimation, detection, amodal segmentation, part detection, classification, and few-shot pose learning. During my path to pursuing PhD, I, with collaboration of others, have developed this approach and extensively studied all components in the pipeline for boarding its application with better accuracy and robustness. The extensive experiments conducted on various datasets demonstrate our approach ability for real-world usage
- …
