Filters








2,060 Hits in 3.0 sec

Self-supervised Learning from a Multi-view Perspective [article]

Yao-Hung Hubert Tsai, Yue Wu, Ruslan Salakhutdinov, Louis-Philippe Morency
2021 arXiv   pre-print
Building from this multi-view perspective, this paper provides an information-theoretical framework to better understand the properties that encourage successful self-supervised learning.  ...  Our theoretical framework paves the way to a larger space of self-supervised learning objective design.  ...  CONCLUSION This work studies both theoretical and empirical perspectives on self-supervised learning.  ... 
arXiv:2006.05576v4 fatcat:slhsl2b7lbcepnw56ehw7oc3c4

Demystifying How Self-Supervised Features Improve Training from Noisy Labels [article]

Hao Cheng, Zhaowei Zhu, Xing Sun, Yang Liu
2021 arXiv   pre-print
The advancement of self-supervised learning (SSL) motivates researchers to apply SSL on other tasks such as learning with noisy labels.  ...  In this paper, we study why and how self-supervised features help networks resist label noise using both theoretical analyses and numerical experiments.  ...  Evaluation of SSL (Self-Supervised Learning): SSL is usually evaluated by two steps: First, use SSL to train an encoder f with only unlabeled data X, then add a linear classifier g following the pre-trained  ... 
arXiv:2110.09022v1 fatcat:aulvwedf45cbthwplwtxeybm3i

Unsupervised Learning via Total Correlation Explanation [article]

Greg Ver Steeg
2017 arXiv   pre-print
Dependence can be characterized using the information-theoretic multivariate mutual information measure called total correlation.  ...  Learning by children and animals occurs effortlessly and largely without obvious supervision.  ...  Demystifying information-theoretic clustering. Mezher, Neda Jahanshad, Talia M. Nir, Xue Hua, Boris A. Gut- In International Conference on Machine Learning, 2014.  ... 
arXiv:1706.08984v1 fatcat:euqu4tgydra5fosirdeovstcru

Learning from psychotherapy for postgraduate supervision

Josie Arnold, Swinburne University of Technology
2008 Journal of University Teaching and Learning Practice  
In his development of an idea that there is in academic writing the self and the researched, the conscious intellectual semiotic and that arising from storytelling, Gregory Ulmer surveys the idea of 'mystories  ...  It develops insights into postgraduate supervision as pedagogy by interrogating the intersection of teaching and learning with some aspects of psychotherapy.  ...  Introduction and Critical Framework The purpose of this paper is to develop insights into postgraduate supervision as pedagogy by interrogating the intersection of teaching and learning with some aspects  ... 
doi:10.53761/1.5.2.5 fatcat:fv6reiaxmbd6xhb6n2vlfoyzau

Graph Augmentation-Free Contrastive Learning for Recommendation [article]

Junliang Yu, Hongzhi Yin, Xin Xia, Lizhen Cui, Quoc Viet Hung Nguyen
2021 arXiv   pre-print
self-supervised manner.  ...  In this paper, we first experimentally demystify that the uniformity of the learned user/item representation distributions on the unit hypersphere is closely related to the recommendation performance.  ...  In our datasets, no features are provided and we mask embeddings learned by DNN to conduct self-supervised learning.  ... 
arXiv:2112.08679v1 fatcat:vg54wz7m2zdfvasu5vjfqm3w5i

Understanding Self-supervised Learning with Dual Deep Networks [article]

Yuandong Tian and Lantao Yu and Xinlei Chen and Surya Ganguli
2021 arXiv   pre-print
We propose a novel theoretical framework to understand contrastive self-supervised learning (SSL) methods that employ dual pairs of deep ReLU networks (e.g., SimCLR).  ...  HLTM) and prove that the hidden neurons of deep ReLU networks can learn the latent variables in HLTM, despite the fact that the network receives no direct supervision from these unobserved latent variables  ...  An analogy between self-supervised and supervised learning: the dual network scenario.  ... 
arXiv:2010.00578v6 fatcat:7l45kjpsn5bv3cxmm5vxqyzqyi

Self-Supervised Metric Learning in Multi-View Data: A Downstream Task Perspective [article]

Shulei Wang
2022 arXiv   pre-print
Self-supervised metric learning has been a successful approach for learning a distance from an unlabeled dataset.  ...  To gain insights into this approach, we develop a statistical framework to theoretically study how self-supervised metric learning can benefit downstream tasks in the context of multi-view data.  ...  Then, we present a new spectral method for self-supervised learning and analyze its theoretical properties in Section 5.  ... 
arXiv:2106.07138v3 fatcat:fmxdfuqygbe5nplsv6fem55ayu

Visual Sensation and Perception Computational Models for Deep Learning: State of the art, Challenges and Prospects [article]

Bing Wei, Yudi Zhao, Kuangrong Hao, Lei Gao
2021 arXiv   pre-print
In this paper, visual perception computational models oriented deep learning are investigated from the biological visual mechanism and computational vision theory systematically.  ...  Visual sensation and perception refers to the process of sensing, organizing, identifying, and interpreting visual information in environmental awareness and understanding.  ...  learning, self-supervised learning, unsupervised learning.  ... 
arXiv:2109.03391v1 fatcat:xtgda2x6azd2laun45tqfj77gi

Various Methods for Object Detection Based on Deep Learning

2019 International journal of recent technology and engineering  
Through this paper demystifies the important role of deep learning supported by CNN for object detection. And the methodology offers additional correct result.  ...  The convolutional natural network determining an object detection task exploitation in deep learning. Object detection is very important in AI as well as in videos using pc vision.  ...  the information used to train is neither classified nor labelled, and it hugely differs from supervised machine learning.  ... 
doi:10.35940/ijrte.c4954.118419 fatcat:5jqoh2tq7zam5a3jfmogx5ga5i

Language learning

2002 Language Teaching  
Modern information technology, especially email, provides an ideal tool to engage learners in genuine written activities.  ...  The theoretical framework, based on the Monitor Model developed by Krashen, is discussed, including some controversial aspects.There were several positive factors arising from the exercise: once the first  ...  The theoretical framework underpinning the project is seen as having important implications for the goals, content, roles and strategies of supervision.  ... 
doi:10.1017/s0261444801221739 fatcat:q5b4ixhfavdgfpteczdw4lfaau

Understanding Undergraduate Research Experiences through the Lens of Problem-based Learning: Implications for Curriculum Translation

Olga Pierrakos, Anna Zilberberg, Robin Anderson
2010 Interdisciplinary Journal of Problem-based Learning  
Problem-based learning (PBL) is a studentcentered pedagogy that offers a strong framework upon which to build a curriculum to teach students essential problem solving skills.  ...  An authentic problem-solving experience, which is highly valued and promoted outside of the classroom yet almost nonexistent in the classroom, is undergraduate research (UR).  ...  Innovative Problem-based Learning Model and Assessment Tools in Undergraduate Engineering Education).  ... 
doi:10.7771/1541-5015.1103 fatcat:sxm6hjaftbcb5n7egf3avd55ou

Contrastive Learning with Hard Negative Samples [article]

Joshua Robinson, Ching-Yao Chuang, Suvrit Sra, Stefanie Jegelka
2021 arXiv   pre-print
We argue that, as with metric learning, contrastive learning of representations benefits from hard negative samples (i.e., points that are difficult to distinguish from an anchor point).  ...  How can you sample good negative examples for contrastive learning?  ...  Demystifying contrastive self-supervised learning: Invariances, augmentations and dataset biases. arXiv:2007.13916, 2020. Florian Schroff, Dmitry Kalenichenko, and James Philbin.  ... 
arXiv:2010.04592v2 fatcat:vbbkxfeoarbhvh7wzuhfufczp4

TSViz: Demystification of Deep Learning Models for Time-Series Analysis

Shoaib Ahmed Siddiqui, Dominique Mercier, Mohsin Munir, Andreas Dengel, Sheraz Ahmed
2019 IEEE Access  
This paper presents a novel framework for demystification of convolutional deep learning models for time-series analysis.  ...  This is a step towards making informed/explainable decisions in the domain of time-series, powered by deep learning.  ...  RELATED WORK Significant efforts have been made to understand the learned features of a model in order to demystify the deep learning black-box.  ... 
doi:10.1109/access.2019.2912823 fatcat:yiaqmn425bdtbhnnt6h4z65baa

Contrastive Attraction and Contrastive Repulsion for Representation Learning [article]

Huangjie Zheng, Xu Chen, Jiangchao Yao, Hongxia Yang, Chunyuan Li, Ya Zhang, Hao Zhang, Ivor Tsang, Jingren Zhou, Mingyuan Zhou
2021 arXiv   pre-print
Contrastive learning (CL) is effective in learning data representations without label supervision, where the encoder needs to contrast each positive sample over multiple negative samples via a one-vs-many  ...  Theoretical analysis reveals the connection between CACR and CL from the perspectives of both positive attraction and negative repulsion and shows the benefits in both efficiency and robustness brought  ...  An efficient framework for learning sentence representations. In International Conference on Learning Representations, 2018.  ... 
arXiv:2105.03746v2 fatcat:nqceaa2aw5hv7mxphhf4jiwuq4

Addressing the eLearning Contradiction: A Collaborative Approach for Developing a Conceptual Framework Learning Object

Colla Jean Macdonald, Emma Stodel, Terrie Thompson, Bill Muirhead, Chris Hinton, Brad Carson, Erin Banit
2005 Interdisciplinary Journal of e-Skills and Lifelong Learning  
In this paper we describe the collaborative process we used to design an online Conceptual Framework Learning Object (C-FLO).  ...  We then illustrate how an interdisciplinary collaborative perspective enhanced both the process and learning outcomes.  ...  (M.A. student) The C-FLO allowed me to take an active and self-directed role in the conceptualisation and development of my conceptual framework.  ... 
doi:10.28945/412 fatcat:ry5r4wgwcvfphoy6k6qruiap4u
« Previous Showing results 1 — 15 out of 2,060 results