Introduction to Semi-Supervised Learning

Zhu, Xiaojin; Goldberg, Andrew B.

doi:10.2200/s00196ed1v01y200906aim006

Cited by 1,401 publications

(1,166 citation statements)

References 72 publications

Supporting

Mentioning

1,073

Contrasting

Unclassified

Order By: Relevance

“…Traditional and state-of-the-art transductive algorithms based on vector space model are [35]: Self-Training [32], Co-Training [3], Expectation Maximization (EM) [16], and Transductive Support Vector Machines (TSVM) [10]. There are also some combinations/variations of these algorithms to perform transductive learning.…”

Section: Related Work Background and Notationsmentioning

confidence: 99%

“…Due to the assumptions and drawbacks of algorithms based on VSM, transductive classification based on networks has been demonstrated to be a useful approach for transductive learning [35]. In such case, the dataset is modeled as a network and the labels of labeled documents are propagated to the unlabeled documents through the network connections.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Term Network Approach for Transductive Classification

Rossi¹,

Rezende²,

Lopes³

2015

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Abstract. Transductive classification is a useful way to classify texts when just few labeled examples are available. Transductive classification algorithms rely on term frequency to directly classify texts represented in vector space model or to build networks and perform label propagation. Related terms tend to belong to the same class and this information can be used to assign relevance scores of terms for classes and consequently the labels of documents. In this paper we propose the use of term networks to model term relations and perform transductive classification. In order to do so, we propose (i) different ways to generate term networks, (ii) how to assign initial relevance scores for terms, (iii) how to propagate the relevance scores among terms, and (iv) how to use the relevance scores of terms in order to classify documents. We demonstrate that transductive classification based on term networks can surpass the accuracies obtained by transductive classification considering texts represented in other types of networks or vector space model, or even the accuracies obtained by inductive classification. We also demonstrated that we can decrease the size of term networks through feature selection while keeping classification accuracy and decreasing computational complexity.

show abstract

Section: Related Work Background and Notationsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Term Network Approach for Transductive Classification

Rossi¹,

Rezende²,

Lopes³

2015

Lecture Notes in Computer Science

View full text Add to dashboard Cite

show abstract

“…The sensor carries large amount of data which when analyzed in a correct way can give better insights that can be useful in taking well informed and precise decisions. Today, deep learning is considered one of the most prominent technique for performing state-of-the-art classification, analysis and predictions (Du & Swamy, 2013;Chapelle et al, 2009;Zhu & Goldberg, 2009;LeCun et al, 2015). This paper presents an IoE based Educational model and discusses the applications, advantages and challenges of using deep learning techniques to develop a learning analytics system by effectively using the IoE big data for taking better and efficient decisions within the educational domain.…”

Section: Introductionmentioning

confidence: 99%

“…This learning can be unsupervised, supervised or semi-supervised (Du & Swamy, 2013;Chapelle et al, 2009;Zhu & Goldberg, 2009;LeCun et al, 2015). In other words we can say that, any deep learning model learns from the experience with minimal external interference.…”

Section: Introductionmentioning

confidence: 99%

Learning analytics for IoE based educational model using deep learning techniques: architecture, challenges and applications

Ahad¹,

Tripathi²,

Agarwal³

2018

Smart Learn. Environ.

View full text Add to dashboard Cite

The new generation teaching-learning pedagogy has created a complete paradigm shift wherein the teaching is no longer confined to giving the content knowledge, rather it fosters the "how, when and why" of applying this knowledge in real world scenarios. By exploiting the advantages of deep learning technology, this pedagogy can be further fine-tuned to develop a repertoire of teaching strategies. This paper presents a secured and agile architecture of an Internet of Everything (IoE) based Educational Model and a Learning Analytics System (LAS) model using the concept of deep learning which can be used to gauge the degree of learning, retention and achievements of the learners and suggests improvements and corrective measures. The paper also puts forward the advantages, applications and challenges of using deep learning techniques for gaining insights from the data generated from the IoE devices within the educational domain for creating such learning analytics systems. Finally a feature wise comparison is provided between the proposed Learning Analytics (LA) based approach and conventional teaching-learning approach in terms of performance parameters like cognition, attention, retention and attainment of learners.

show abstract

“…In a recent paper, Singh et al (2008) established that if the complexity of the distributions under consideration is too high to be understood using labeled data points, but is small enough to be understood using unlabeled data points, using a finite sample analysis in SSL can improve the performance of a supervised learning task. There have been many successful practical SSL algorithms generated as summarized in Chapelle et al (2006), Sindhwani (2005), Zhu (2005), Xu et al (2010) and Zhu and Goldberg (2009).…”

Section: Introductionmentioning

confidence: 99%

Study on semi-supervised local constant regression estimation

Seok¹

2012

Journal of the Korean Data and Information Science Society

View full text Add to dashboard Cite

Many different semi-supervised learning algorithms have been proposed for use with unlabeled data. However, most of them focus on classification problems. In this paper, we propose a semi-supervised regression algorithm called the semi-supervised local constant estimator (SSLCE), based on the local constant estimator (LCE), and reveal the asymptotic properties of SSLCE. We also show that the SSLCE has a faster convergence rate than that of the LCE when a well chosen weighting factor is employed. Our experiment with synthetic data shows that the SSLCE can improve performance with unlabeled data, and we recommend its use with the proper size of unlabeled data.

show abstract

Introduction to Semi-Supervised Learning

Cited by 1,401 publications

References 72 publications

Term Network Approach for Transductive Classification

Term Network Approach for Transductive Classification

Learning analytics for IoE based educational model using deep learning techniques: architecture, challenges and applications

Study on semi-supervised local constant regression estimation

Contact Info

Product

Resources

About