502 Hits in 4.3 sec

Optimal 1-NN prototypes for pathological geometries

Ilia Sucholutsky, Matthias Schonlau
2021 PeerJ Computer Science  
The number and distribution of prototypes required for the classifier to match its original performance is intimately related to the geometry of the training data.  ...  Finally, we show that a parametric prototype generation method that normally cannot solve this pathological setting can actually find optimal prototypes when combined with the results of our theoretical  ...  Certain pathological geometries can result in especially poor performance of these heuristic algorithms.  ... 
doi:10.7717/peerj-cs.464 pmid:33954242 pmcid:PMC8049135 fatcat:ny32ylc7zrcqxc2yven3hqk2xq

Optimal 1-NN Prototypes for Pathological Geometries [article]

Ilia Sucholutsky, Matthias Schonlau
2020 arXiv   pre-print
The number and distribution of prototypes required for the classifier to match its original performance is intimately related to the geometry of the training data.  ...  As a result, it is often difficult to find the optimal prototypes for a given dataset, and heuristic algorithms are used instead.  ...  Certain pathological geometries can result in especially poor performance of these heuristic algorithms.  ... 
arXiv:2011.00228v1 fatcat:q5udrxke5fcz5hr7oaszhb7xam

Automatic detection of the view position of chest radiographs

Thomas M. Lehmann, Mark O. Gueld, Daniel Keysers, Henning Schubert, Andrea Wenning, Berthold B. Wein, Milan Sonka, J. Michael Fitzpatrick
2003 Medical Imaging 2003: Image Processing  
In the second step, the normalized cross correlation function at the optimal displacement is used for 5-nearest-neighbor classification.  ...  ., these results show that the determination of the view position of chest radiographs can be fully automated and substantially simplified if the correlation function is used directly for 5-NN classification  ...  Referring to only 5 size of leaving-one-out experiments 5 prototypes 430 prototypes 2 templates feature image 1-NN 5-NN 1-NN 1-NN 5-NN 1-NN 64 x 64 99.09 % 98.66 % 90.62 % 98.66 % 98.77  ... 
doi:10.1117/12.481404 dblp:conf/miip/LehmannGKSWW03 fatcat:4scbyklhmjg6pj5p2w3snez7rq

Large margin linear discriminative visualization by Matrix Relevance Learning

Michael Biehl, Kerstin Bunte, Frank-Michael Schleif, Petra Schneider, Thomas Villmann
2012 The 2012 International Joint Conference on Neural Networks (IJCNN)  
This prototype-based, supervised learning scheme parameterizes an adaptive distance measure in terms of a matrix of relevance factors.  ...  By means of a few benchmark problems, we demonstrate that the training process yields low rank matrices which can be used efficiently for the discriminative visualization of labeled data.  ...  More importantly, a different cost function is optimized which appears to be less sensitive to the specific cluster geometry in many cases.  ... 
doi:10.1109/ijcnn.2012.6252627 dblp:conf/ijcnn/BiehlBSSV12 fatcat:magnb5gvbrfavhuedmmjv7vtqq

Photoacoustic spectroscopy of ovarian normal, benign, and malignant tissues: a pilot study

Sudha D. Kamath, Satadru Ray, Krishna K. Mahato
2011 Journal of Biomedical Optics  
Bhat for providing some of the ovarian tissues used in the study.  ...  Satyamoothy, Professor and Director, Manipal Life Sciences Centre for providing the necessary facilities to carry out this study. The authors are also thankful to Dr. Rani A.  ...  For classification, a prototype sample is computed from the reference (calibration) set and a given test sample is classified as belonging to the class of the closest prototype.  ... 
doi:10.1117/1.3583573 pmid:21721822 fatcat:ccnnq6pzfzgb5hkmnwq5ml3vgm

Design of a reconfigurable autoencoder neural network for detector front-end ASICs

Jim Hirschauer
2021 Zenodo  
layer Encoder NN Optimization of dimensions shown next Encoder NN architecture optimization or or 5×5 3×3 4×4×3 8×8 2-10 Geometry mapping # of conv2D filters conv2D kernel size ➔  ...  physical radiation patterns NN encoding layers Map to a regular geometry 2d 3d or Scan filters to extract features • A fully-connected NN layer achieves further reduction, exploiting  ... 
doi:10.5281/zenodo.4640887 fatcat:5ybluuluezdzfplm6o5zspxw7a

Determining the View of Chest Radiographs

Thomas M. Lehmann, O. G�ld, Daniel Keysers, Henning Schubert, Michael Kohnen, Berthold B. Wein
2003 Journal of digital imaging  
Remaining errors are caused by image altering pathologies, metal artifacts, or other interferences with routine conditions.  ...  For comparison to existing approaches, subsets of 430 and 5 training images are also considered.  ...  The experiments based on 1,867 and 430 references were executed using a 5-NN classifier, while those referring to 5 reference images were based on a 1-NN classifier.  ... 
doi:10.1007/s10278-003-1655-x pmid:14669063 pmcid:PMC3045251 fatcat:dmv735bbzjdg7brzdn3chmhq3q

Universal consistency and rates of convergence of multiclass prototype algorithms in metric spaces [article]

László Györfi, Roi Weiss
2021 arXiv   pre-print
While obtaining rates for Proto-NN is left open, we show that a second prototype rule that hybridizes between k-NN and Proto-NN achieves the same rates as k-NN while enjoying similar computational advantages  ...  We study universal consistency and convergence rates of simple nearest-neighbor prototype rules for the problem of multiclass classification in metric paces.  ...  We thank the editor and referees for carefully reading the manuscript and for the suggested improvements.  ... 
arXiv:2010.00636v2 fatcat:wjs4oatvcbarvd7oxn3kebwhiq

Distributionally Robust Weighted k-Nearest Neighbors [article]

Shixiang Zhu and Liyan Xie and Minghe Zhang and Rui Gao and Yao Xie
2022 arXiv   pre-print
In this paper, we study a minimax distributionally robust formulation of weighted k-nearest neighbors, which aims to find the optimal weighted k-NN classifiers that hedge against feature uncertainties.  ...  We develop an algorithm, , that efficiently solves this functional optimization problem and features in assigning minimax optimal weights to training samples when performing classification.  ...  , P * M ) leads to the same optimal value for formulation (6) as (π * ; P * 1 , . . . , P * M ) for formulation (8).  ... 
arXiv:2006.04004v5 fatcat:5mr44gim6vbivcs7sohvpiug5e

Improving FEM crash simulation accuracy through local thickness estimation based on CAD data

Vânio Ferreira, Luís Paulo Santos, Markus Franzen, Omar O. Ghouati, Ricardo Simoes
2014 Advances in Engineering Software  
the benefits of computer simulations in engineering design by enabling zero-prototyping and thus reducing product development costs.  ...  Finite Element Methods (FEM) have been a standard tool for a long time, particularly for modeling and simulating thermoplastic parts under service conditions [12, 16] .  ...  Acknowledgements The authors would like to thank the FCT -Fundação para a Ciência e a Tecnologia (Portuguese Foundation for Science and Technology) through projects PEst-C/CTM/LA0025/2013 and PEst-OE/EEI  ... 
doi:10.1016/j.advengsoft.2014.02.003 fatcat:r2rt5sdr4zgupmgmn5p4mygbqe


2004 Biomedical Engineering Applications Basis and Communications  
Moreover, the feature vector based on fractal geometry and wavelet transform can provide good discriminant ability for ultrasonic liver images under study.  ...  In this study, we evaluate the accuracy of classifiers for classification of ultrasonic liver tissues.  ...  >(A"-*) > (') k Neighbor (fr-NN) Classifier [47] [48] : In a &-NN classifier, each class is represented by a set of prototype vectors.  ... 
doi:10.4015/s1016237204000104 fatcat:dplc3hnv4bhr3b7vw6hwb43w5u

Three-dimensional virtual histology of the human hippocampus based on phase-contrast computed tomography

Marina Eckermann, Bernhard Schmitzer, Franziska van der Meer, Jonas Franz, Ove Hansen, Christine Stadelmann, Tim Salditt
2021 Proceedings of the National Academy of Sciences of the United States of America  
Accordingly, we find that the prototypical transformation between a structure representing healthy granule cells and the pathological state involves a decrease in the volume of granule cell nuclei, as  ...  learning, representation of structural properties in a feature space, and classification based on the theory of optimal transport.  ...  of interest (ROI) scanned at high magnification based on cone beam geometry.  ... 
doi:10.1073/pnas.2113835118 pmid:34819378 pmcid:PMC8640721 fatcat:jmmpgvpbv5gevdtxp5c6cfzab4

A virtual sizing tool for mitral valve annuloplasty

Manuel K. Rausch, Alexander M. Zöllner, Martin Genet, Brian Baillargeon, Wolfgang Bothe, E. Kuhl
2016 International Journal for Numerical Methods in Biomedical Engineering  
a 24mm ring, 1.9N for a 28mm ring, and 0.8N for a 32mm ring.  ...  Here we prototype a virtual sizing tool to quantify changes in annular dimensions, surgically-induced tissue strains, mitral annular stretches, and suture forces in response to mitral annuloplasty.  ...  Our overall goal is to prototype a virtual surgical tool for the optimal design and selection of cardiovascular devices using predictive computational simulation.  ... 
doi:10.1002/cnm.2788 pmid:27028496 pmcid:PMC5289896 fatcat:ysclmqkgnbf55lq6y2h4oh6co4

Platform AF: Membrane Engineering

2008 Biophysical Journal  
The results allow for an optimization of the molecular architecture.  ...  $ 6 nN for every extra pair of ÀCH 2 groups present in the chain along the series DMPC-DSPC.  ... 
doi:10.1016/s0006-3495(08)79058-1 fatcat:sjixny7y7vf3tal6aeemwi62mu

A uniaxial bioMEMS device for imaging single cell response during quantitative force-displacement measurements

David B. Serrell, Jera Law, Andrew J. Slifka, Roop L. Mahajan, Dudley S. Finch
2008 Biomedical microdevices  
A microfabricated device has been developed for imaging of a single, adherent cell while quantifying force under an applied displacement.  ...  Nick Barbosa III for the SEM images. This work is a contribution of the United States Department of Commerce and is not subject to copyright in the United State.  ...  Calibration curves converge on the theoretical value (b) Two other cells were measured using this device, which resulted in values of k 1 =173.24 nN/µm, k 2 =95.90 nN/µm, τ=38.37 s and k 1 =192.00 nN/µm  ... 
doi:10.1007/s10544-008-9202-7 pmid:18648937 fatcat:hpxbm4b3tzcbfkio52za6jhbqe
« Previous Showing results 1 — 15 out of 502 results