Feature combination using boosting

Yin, Xu-Cheng; Liu, Changping; Han, Zhi

doi:10.1016/j.patrec.2005.03.029

Cited by 54 publications

(25 citation statements)

References 14 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…The traditional boosting produces only one component weak classifier at each iteration. By contrast, at each round of our extension of the boosting procedures, several weak classifiers are trained on samples of each feature, and then combined into a single one (a middle final classifier) [38]:…”

Section: Boostingmentioning

confidence: 99%

See 1 more Smart Citation

Cost-sensitive learning in social image tagging: review, new ideas and evaluation

Lew

2012

Int J Multimed Info Retr

View full text Add to dashboard Cite

Visual concept learning typically requires a set of expert labeled, manual training images. However, acquiring a sufficient number of reliable annotations can be timeconsuming or impractical. Therefore, in many situations it is preferable to perform unsupervised learning on user contributed tags from abundant sources such as social Internet communities and websites. Cost-sensitive learning is a natural approach toward unsupervised visual concept learning because it fundamentally optimizes the learning system accuracy regarding the cost of an error. This paper reviews the problem of cost-sensitive unsupervised learning of visual concepts from social images, presents the new ideas, and gives a comparative evaluation of representative approaches from the research literature.

show abstract

Section: Boostingmentioning

confidence: 99%

“…Therefore, a second heuristic idea is to formulate it inside the exponent of the cost function: (38) Now the second order Taylor approximation we want to optimize is defined as follows:…”

Section: Importance Weighted Gentleboostmentioning

confidence: 99%

Cost-sensitive learning in social image tagging: review, new ideas and evaluation

Lew

2012

Int J Multimed Info Retr

View full text Add to dashboard Cite

show abstract

“…Multiple classifier combination has been intensively studied for many years [23][24][25][26]. It is shown that combination of multiple neural networks can achieve higher performance over the best individual one.…”

Section: Integrated Prediction For Exchange Ratementioning

confidence: 99%

Exchange rate prediction with non-numerical information

Wang

Hao

Yin

et al. 2010

Neural Comput & Applic

Self Cite

View full text Add to dashboard Cite

Exchange rate prediction is an important yet challenging problem in financial time series analysis. Although the historical exchange rates can provide valuable information, other factors will also affect the prediction significantly. These factors could be numerical or non-numerical ones, which are related to politics, economics, military, or even market psychology. Previous automatic exchange rate prediction merely considers numerical data (or simply the historical rates) for predicting the next day value. In this paper, we show how to utilize and combine many related factors, both numerical and non-numerical factors, for exchange rate prediction. With an example on forecasting exchange rate between US dollar and Japanese yen, we investigate how to exploit the information from non-numerical factors. We then engage a novel integrated approach which successfully combines information obtained from both numerical and nonnumerical factors. We show how to quantify the nonnumerical fundamental information, provide details steps on how to construct single predictors on different kinds of information separately, and finally describe how to integrate these separate predictors. Experimental results showed that our method can achieve the Directional Symmetry (DS) accuracy of 86.96%, which is much higher than only exploiting numerical information.

show abstract

“…Based on a given classifier set, the combination methods can be categorized according to the level of classifier outputs: abstract level (class label), rank level (rank order), and measurement level (class scores) [12,18]. Yin et al analyze multiple classifier systems from the perspective of feature combination [15]. And most popular systems are with multiple classifiers for character recognition [4,12,15].…”

Section: Introductionmentioning

confidence: 99%

“…Yin et al analyze multiple classifier systems from the perspective of feature combination [15]. And most popular systems are with multiple classifiers for character recognition [4,12,15].…”

Section: Introductionmentioning

confidence: 99%