Classification with learning k-nearest neighbors

Laaksonen, Jorma; Oja, Erkki

doi:10.1109/icnn.1996.549118

Cited by 121 publications

(82 citation statements)

References 3 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Friedman [19] proposed an interesting way of adapting the metric based on a tree-structure interactive partitioning of the data. Laaksonen and Oja [4] proposed to improve the k-NN reference vectors using LVQ techniques. Atkenson, Moor and Schaal [10] discuss locally weighted regression techniques, minimal distance methods with various metric and kernel functions applied to approximation problems.…”

Section: Summary and Discussionmentioning

confidence: 99%

“…4, therefore smaller weights of the norm mean that the network operates in an almost linear regime. Regularization methods add penalty terms to the error function forcing the weights to become small and thus smoothing the network approximation to the training data.…”

Section: A Framework For Minimal Distance Methodsmentioning

confidence: 99%

“…Various rules for moving centers D are used: moving only the nearest neighbor, moving all k neighbors by the same amount, using distance-dependent η etc. [4]. One can also optimize a subset of vectors, for example only those that are close to the center of clusters.…”

Section: Parameterization Of Reference Vectorsmentioning

confidence: 99%

“…If the new training vector falls into the range of a sphere of a wrong class the radius of this sphere is shrinked to leave the vector outside of the sphere. Positions of the spheres are not optimized in the RCE algorithm (this would lead in the direction of LVQ algorithms [4]), but voting methods for the committees of classifiers were used with success [8]. The number of radiuses r i may be reduced by using only a few independent values in selected input space areas.…”

Section: R-nn Algorithmmentioning

confidence: 99%

“…One problem is the computational complexity of the actual classification, demanding for n reference vectors calculation of ∼ n 2 distances and finding the k smallest distances. Although Laaksonen and Oja [4] claim that "For realistic pattern space dimensions, it is hard to find any variation of the rule that would be significantly lighter than the brute force method" various hierarchical schemes of partitioning the data space or hierarchical clusterization schemes are quite effective. The search for the nearest neighbors is easily paralelizable and the training time (selection of optimal k) is relatively short.…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

Minimal distance neural methods

Dich

Grudziński

Diercksen

1998 IEEE International Joint Conference on Neural Networks Proceedings. IEEE World Congress on Computational Intelligence (Cat

View full text Add to dashboard Cite

Section: Summary and Discussionmentioning

confidence: 99%

Section: A Framework For Minimal Distance Methodsmentioning

confidence: 99%

Section: Parameterization Of Reference Vectorsmentioning

confidence: 99%

Section: R-nn Algorithmmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Minimal distance neural methods

Dich

Grudziński

Diercksen

1998 IEEE International Joint Conference on Neural Networks Proceedings. IEEE World Congress on Computational Intelligence (Cat

View full text Add to dashboard Cite

NCA‐based hybrid convolutional neural network model for classification of cervical cancer on gauss‐enhanced pap‐smear images

Bingol

2022

Int J Imaging Syst Tech

View full text Add to dashboard Cite

Cervical cancer is a very serious disease that deeply affects women's lives, often resulting in death. This type of cancer, which is very common in women, is diagnosed at an early stage and is of vital importance for the success of the treatment. Pap-smear tests are used by physicians as the primary diagnostic tool to diagnose the disease. In this study, a hybrid deep model is proposed to classify pap-smear images to detect cervical cancer. In addition, the Gaussian method was applied to improve the images in the original dataset. Feature maps were taken from both the original dataset and the Gaussian-enhanced dataset in the built hybrid architecture, which used Darknet53 and Mobilenetv2 models as the base. After these obtained feature maps were combined, useless features were extracted and the number of features was reduced by using the Neighborhood Component Analysis (NCA) dimension reduction method. Finally, this optimized feature map was classified into different classifiers. As a result of the experimental studies, it was determined that the proposed hybrid model performed better when compared to other studies in the literature and the accuracy rate was 98.90% in the Support Vector Machines (SVM) classifier.

show abstract

Machine learning applications in vadose zone hydrology: A review

Li,

Nieber,

Kumar

2024

Vadose Zone Journal

View full text Add to dashboard Cite

Machine learning (ML) has been broadly applied for vadose zone applications in recent years. This article provides a comprehensive review of such developments. ML applications for variables corresponding to different complex vadose zone processes are summarized mostly in a prediction context. By analyzing and assessing these applications, we discovered extensive usages of classic ML models with relatively limited applications of deep learning (DL) approaches in general. We also recognized a lack of benchmark datasets for soil property research as well as limited integration of physics‐based vadose zone principles into the ML approaches. To facilitate this interdisciplinary research of ML in vadose zone characterization and processes, a paradigm of knowledge‐guided machine learning is suggested along with other data‐driven and ML model‐based research suggestions to advance future research.

show abstract

Classification with learning k-nearest neighbors

Cited by 121 publications

References 3 publications

Minimal distance neural methods

Minimal distance neural methods

NCA‐based hybrid convolutional neural network model for classification of cervical cancer on gauss‐enhanced pap‐smear images

Machine learning applications in vadose zone hydrology: A review

Contact Info

Product

Resources

About