138,934 Hits in 2.8 sec

Addressing Class Imbalance in Classification Problems of Noisy Signals by using Fourier Transform Surrogates [article]

Justus T. C. Schwabedal, John C. Snyder, Ayse Cakmak, Shamim Nemati, Gari D. Clifford
2019 arXiv   pre-print
We show how such conditional confusion matrices can qualitatively explain the performance of surrogates in class balancing.  ...  The visualization is based on the response of inferred class probabilities under replacement of short data segments by partial surrogates.  ...  We computed the number of repetitions n c of under-represented class c necessary to match the number of the most frequent class.  ... 
arXiv:1806.08675v2 fatcat:zboz4iwz5faiplmefbn3ac7qoi

Photochemical changes in water accommodated fractions of MC252 and surrogate oil created during solar exposure as determined by FT-ICR MS

Pamela P. Vaughan, Tashiema Wilson, Rebecca Kamerman, Melissa E. Hagy, Amy McKenna, Huan Chen, Wade H. Jeffrey
2016 Marine Pollution Bulletin  
Photochemical alterations resulted in differences in compound class distributions. In general, surrogate oil was photo-oxidized across a wider carbon number range compared to MC252.  ...  To determine effects of photochemical weathering of petroleum, surrogate and Macondo (MC252) crude oils were exposed to solar radiation during the formation of Water Accommodated Fractions (WAF) in sterile  ...  Differences in heteroatom class distribution for full sun incubated WAFs were observed between surrogate and MC252.  ... 
doi:10.1016/j.marpolbul.2016.01.012 pmid:26774346 fatcat:5zbn2wctbja4fe53jbpu3432ly

Deep Unsupervised Similarity Learning Using Partially Ordered Sets

Miguel A. Bautista, Artsiom Sanakoyeu, Bjorn Ommer
2017 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR)  
To overcome this problem, we use local estimates of reliable (dis-)similarities to initially group samples into compact surrogate classes and use local partial orders of samples to classes to link classes  ...  Adopting a strategy of self-supervision, a CNN is trained to optimally represent samples in a mutually consistent manner while updating the classes.  ...  In Eq. (4) C ′ is the number of surrogate classes in the batch, σ is the standard deviation of the current assignment of samples to surrogate classes, and γ is the margin between surrogate classes.  ... 
doi:10.1109/cvpr.2017.208 dblp:conf/cvpr/BautistaSO17 fatcat:jag3zwtzpva7fkm6uulw5mp4ku

Deep Unsupervised Similarity Learning using Partially Ordered Sets [article]

Miguel A Bautista, Artsiom Sanakoyeu, Björn Ommer
2017 arXiv   pre-print
To overcome this problem, we use local estimates of reliable (dis-)similarities to initially group samples into compact surrogate classes and use local partial orders of samples to classes to link classes  ...  Adopting a strategy of self-supervision, a CNN is trained to optimally represent samples in a mutually consistent manner while updating the classes.  ...  Aknowledgments: This research has been funded in part by the Heidelberg Academy of Sciences. We are grateful to the NVIDIA corporation for donating a Titan X GPU.  ... 
arXiv:1704.02268v3 fatcat:bmisdr5plzgg7hsdhuxlgljbeu

Machine Learning-Based Identification Strategy of Fuel Surrogates for the CFD Simulation of Stratified Operations in Low Temperature Combustion Modes

Valerio Mariani, Leonardo Pulga, Gian Marco Bianchi, Stefania Falfari, Claudio Forte
2021 Energies  
Many researchers in industry and academia are showing an increasing interest in the definition of fuel surrogates for Computational Fluid Dynamics simulation applications.  ...  The developed methodology has been applied on different real gasolines from literature in order to identify their optima surrogates.  ...  Figure 8 . 8 EPIONA class-by-class and overall composition compatibility of optima and bound surrogates of the four validation gasolines (A, B, C, D).  ... 
doi:10.3390/en14154623 fatcat:6vpswj4y2jfihc3l6lszoojdy4

Crowdsourcing of network data

Ding Wang, Prakash Mandayam Comar, Pang-Ning Tan
2016 2016 International Joint Conference on Neural Networks (IJCNN)  
To overcome this limitation, we present a novel framework called surrogate learning to transform the network data into a new representation (i.e., images) so that the labeling task can be completed even  ...  We analyze the reconstruction error of the transformation and use the theoretical insights to provide guidance on how to develop an effective surrogate learning approach for any given network and source  ...  (pixels), and c s is the number of classes.  ... 
doi:10.1109/ijcnn.2016.7727472 dblp:conf/ijcnn/WangCT16 fatcat:7befyyrarrcodk5nialrr5kl3q

Surrogate data: A novel approach to object detection

Zbisław Tabor
2010 International Journal of Applied Mathematics and Computer Science  
The method consists in generating surrogate data for an analyzed image.  ...  Surrogate data: A novel approach to object detection In the present study a novel method is introduced to detect meaningful regions of a gray-level noisy images of binary structures.  ...  that the gray-level intensities at all path pixels do not exceed I C exists in surrogate data.  ... 
doi:10.2478/v10006-010-0040-4 fatcat:2qhdiezykvdztkft5j5xzd5wsq

Simple strategies for semi-supervised feature selection

Konstantinos Sechidis, Gavin Brown
2017 Machine Learning  
These minimalist, seemingly naive, approaches have not previously been studied in depth.  ...  In the context of semi-supervised feature selection, we tackle exactly this-how much we can gain from two simple classifier-independent strategies.  ...  2 -G(X; Y0) Surrogate 3 -G(X; Y1) (b) Testing: labels missing at random class dependent (MAR-C) When the labels are MAR-C we prove with the following theorem that we have the same valid tests as in  ... 
doi:10.1007/s10994-017-5648-2 pmid:31983804 pmcid:PMC6954040 fatcat:ze4ljb4xlng7zdw5btvdcz52ue

Boosting Nearest Neighbors for the Efficient Estimation of Posteriors [chapter]

Roberto D'Ambrosio, Richard Nock, Wafa Bel Haj Ali, Frank Nielsen, Michel Barlaud
2012 Lecture Notes in Computer Science  
In this paper, we analyze, in the light of this problem, a recent algorithm, unn, which leverages nearest neighbors while minimizing a convex loss. Our contribution is threefold.  ...  It is an admitted fact that mainstream boosting algorithms like AdaBoost do not perform well to estimate class conditional probabilities.  ...  In a multiclass multilabel setting, we may compute the total surrogate risk over all classes as: ε ψ S (H) . = 1 C C c=1 ε ψ S (h c , c) , (15) where H is the set of all C classifiers h 1 , h 2 , ...,  ... 
doi:10.1007/978-3-642-33460-3_26 fatcat:neiqfaepz5e5hlnywnbs4zusiu

A surrogate loss function for optimization of F_β score in binary classification with imbalanced data [article]

Namgil Lee, Heejung Yang, Hojin Yoo
2021 arXiv   pre-print
Through numerical experiments using ResNets and benchmark image data sets, it is demonstrated that the proposed surrogate F_β loss function is effective for optimizing F_β scores under class imbalances  ...  in binary classification tasks compared with other loss functions.  ...  (a) Surrogate F β loss v.s. F 0.5 score (b) Surrogate F β loss v.s. F 1.0 score (c) Surrogate F β loss v.s. F 2.0 score (d) Surrogate F β loss v.s.  ... 
arXiv:2104.01459v1 fatcat:jspig4w6bfhmpmwrw47t45we3q

LIMEtree: Interactively Customisable Explanations Based on Local Surrogate Multi-output Regression Trees [article]

Kacper Sokol, Peter Flach
2020 arXiv   pre-print
In this work we introduce a model-agnostic and post-hoc local explainability technique for black-box predictions called LIMEtree, which employs surrogate multi-output regression trees.  ...  While other methods may give an illusion of customisability by wrapping, otherwise static, explanations in an interactive interface, our explanations are truly interactive, in the sense of allowing the  ...  The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.  ... 
arXiv:2005.01427v1 fatcat:xyvqoy5fivh7darsmbo54x5tdu

The Lovasz-Softmax Loss: A Tractable Surrogate for the Optimization of the Intersection-Over-Union Measure in Neural Networks

Maxim Berman, Amal Rannen Triki, Matthew B. Blaschko
2018 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition  
We present a method for direct optimization of the mean intersection-over-union loss in neural networks, in the context of semantic image segmentation, based on the convex Lovász extension of submodular  ...  appropriate relevance to small objects, and appropriate counting of false negatives, in comparison to per-pixel losses.  ...  common in semantic segmentation, we average the class-specific surrogates; hence we define the Lovász-Softmax loss as loss(f ) = 1 |C| cC ∆ Jc (m(c)) (13) which is piecewise linear in f , the normalized  ... 
doi:10.1109/cvpr.2018.00464 dblp:conf/cvpr/BermanTB18 fatcat:t4trtsaxtvg33g6seavjasuat4

Boosting Black-Box Attack with Partially Transferred Conditional Adversarial Distribution [article]

Yan Feng, Baoyuan Wu, Yanbo Fan, Li Liu, Zhifeng Li, Shutao Xia
2021 arXiv   pre-print
However, due to the possible differences on model architectures and training datasets between surrogate and target models, dubbed "surrogate biases", the contribution of adversarial transferability to  ...  To tackle this issue, we innovatively propose a black-box attack method by developing a novel mechanism of adversarial transferability, which is robust to the surrogate biases.  ...  We pre-train the c-Glow model with four surrogate models on ImageNet, as described in Sec. 4.2.2.  ... 
arXiv:2006.08538v4 fatcat:pila4sc75few3picgf44csm46m

Entities and Surrogates in Knowledge Representation [chapter]

Michel Chein
2010 Lecture Notes in Computer Science  
) there is at most one surrogate in any equivalence class of R(F) and (ii) for any C in C there is no homomorphism from C to R(F).  ...  For any constant c in const(K), we note e c the single individual surrogate belonging to the equivalence class of c.  ... 
doi:10.1007/978-3-642-14197-3_1 fatcat:mh6hzanlrjg45if5f62sgwrsle

Two-temperature logistic regression based on the Tsallis divergence [article]

Ehsan Amid, Manfred K. Warmuth, Sriram Srinivasan
2019 arXiv   pre-print
The surrogate loss of an example with linear activation vector a and class c has the form -log_t_1_t_2 (a_c - G_t_2(a)) where the two temperatures t_1 and t_2"temper" the log and , respectively, and G_t  ...  The algorithm has one weight vector per class and the surrogate loss is a function of the linear activations (one per class).  ...  The algorithm has one weight vector per class and the surrogate loss is a function of the linear activations (one per class).  ... 
arXiv:1705.07210v2 fatcat:ykopju5h3ngtli2axqzqxqn4le
« Previous Showing results 1 — 15 out of 138,934 results