2013
DOI: 10.1007/978-3-319-03536-9_5
|View full text |Cite
|
Sign up to set email alerts
|

On the Choice of Kernel and Labelled Data in Semi-supervised Learning Methods

Abstract: Abstract. Semi-supervised learning methods constitute a category of machine learning methods which use labelled points together with unlabelled data to tune the classifier. The main idea of the semi-supervised methods is based on an assumption that the classification function should change smoothly over a similarity graph, which represents relations among data points. This idea can be expressed using kernels on graphs such as graph Laplacian. Different semi-supervised learning methods have different kernels wh… Show more

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
2
1
1
1

Citation Types

0
18
0

Year Published

2014
2014
2022
2022

Publication Types

Select...
5
2

Relationship

2
5

Authors

Journals

citations
Cited by 16 publications
(18 citation statements)
references
References 15 publications
0
18
0
Order By: Relevance
“…As it is thoroughly commented in [18], SSL methods show to perform remarkably well when only a few labelled points are known. Let us now apply this SSL scheme to our mapping problem, seeking for a more relevant and easier to determine mapping between the nodes of the graph and the amplitudes set of the corresponding time series.…”
Section: Duality and Graph-based Semi-supervised Learningmentioning
confidence: 90%
“…As it is thoroughly commented in [18], SSL methods show to perform remarkably well when only a few labelled points are known. Let us now apply this SSL scheme to our mapping problem, seeking for a more relevant and easier to determine mapping between the nodes of the graph and the amplitudes set of the corresponding time series.…”
Section: Duality and Graph-based Semi-supervised Learningmentioning
confidence: 90%
“…For example, Facebook (which is an undirected social network) used Personalized PageRank for friend recommendation [5]. The social network Twitter is directed, but Twitter's friend recommendation algorithm (Who to Follow) [16] uses an algorithm called personalized SALSA [19,6], which first converts the directed network into an expanded undirected graph 3 , and then computes PPR on this new graph. Random walks have also been used for collaborative filtering by the YouTube team [7] (on the undirected user-item bipartite graph), to predict future items a user will view.…”
Section: Introductionmentioning
confidence: 99%
“…which looks like an interesting inequality for Personalized PageRank. Due to symmetry, K modifPPR ij = K modifPPR ji , and we obtain an independent proof of the following identity for Personalized PageRank [2]:…”
Section: Then Considermentioning
confidence: 83%
“…Informally, given a weighted graph G, a similarity measure on the set of its vertices V (G) is a function κ : V (G) × V (G) → R that characterizes similarity (or affinity, or closeness) between the vertices of G in a meaningful manner and thus is intuitively and practically adequate for empirical applications [2,18,24,33].…”
Section: Definitions and Preliminariesmentioning
confidence: 99%