Speeding up the testing and training time for the support vector machines with minimal effect on the performance

Ghaffari, Hamid Reza

doi:10.1007/s11227-021-03729-0

Cited by 2 publications

(4 citation statements)

References 25 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…These observations and their interpretations substantiate the effectiveness of prioritizing selection from non-overlapping regions in preserving class boundary information. Adopting this selection strategy aligns with Gaffari’s method [ 17 ], which aims at reducing the complexity of the decision boundary through the removal of harmful points (i.e., those with other classes dominating their neighborhoods). However, it contrasted with other methods such as CBCH [ 15 ] and BPLSH [ 16 ], where boundaries are identified based on impure-class neighborhoods.…”

Section: Discussionmentioning

confidence: 99%

“…The Jaccard index is high (0.93) primarily because CBCH retains instances in areas of overlap between the two classes, and most of those are considered support vectors by the SVM algorithm. In addition, the proposed methods are compared to FIFDR [ 13 ] and BPLSH [ 16 ] according to reported results in the literature [ 17 ]. The compared results are presented in Table 3 .…”

Section: Methodsmentioning

confidence: 99%

“…Our findings were compared to those of FIFDR [ 13 ], Gaffari’s method [ 17 ] and Shell Extraction (SE) [ 14 ] according to the literature [ 17 ]. Table 11 summarizes the comparison.…”

Section: Methodsmentioning

confidence: 99%

“…In his approach, Gaffari [ 17 ] proposed a method aiming to reduce both the data set size and boundary complexity. Based on the neighborhood information of each data point, harmful data points are removed.…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Data reduction for SVM training using density-based border identification

Shalaby,

Farouk,

Khater

2024

PLoS ONE

View full text Add to dashboard Cite

Numerous classification and regression problems have extensively used Support Vector Machines (SVMs). However, the SVM approach is less practical for large datasets because of its processing cost. This is primarily due to the requirement of optimizing a quadratic programming problem to determine the decision boundary during training. As a result, methods for selecting data instances that have a better likelihood of being chosen as support vectors by the SVM algorithm have been developed to help minimize the bulk of training data. This paper presents a density-based method, called Density-based Border Identification (DBI), in addition to four different variations of the method, for the lessening of the SVM training data through the extraction of a layer of border instances. For higher-dimensional datasets, the extraction is performed on lower-dimensional embeddings obtained by Uniform Manifold Approximation and Projection (UMAP), and the resulting subset can be repetitively used for SVM training in higher dimensions. Experimental findings on different datasets, such as Banana, USPS, and Adult9a, have shown that the best-performing variations of the proposed method effectively reduced the size of the training data and achieved acceptable training and prediction speedups while maintaining an adequate classification accuracy compared to training on the original dataset. These results, as well as comparisons to a selection of related state-of-the-art methods from the literature, such as Border Point extraction based on Locality-Sensitive Hashing (BPLSH), Clustering-Based Convex Hull (CBCH), and Shell Extraction (SE), suggest that our proposed methods are effective and potentially useful.

show abstract

Section: Discussionmentioning

confidence: 99%

Section: Methodsmentioning

confidence: 99%

“…Our findings were compared to those of FIFDR [ 13 ], Gaffari’s method [ 17 ] and Shell Extraction (SE) [ 14 ] according to the literature [ 17 ]. Table 11 summarizes the comparison.…”

Section: Methodsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations