Filters








470,791 Hits in 2.8 sec

Learning Low-Density Separators [article]

Shai Ben-David, Tyler Lu, David Pal, Miroslava Sotakova
2009 arXiv   pre-print
We define a novel, basic, unsupervised learning problem - learning the lowest density homogeneous hyperplane separator of an unknown probability distribution.  ...  We propose two natural learning paradigms and prove that, on input unlabeled random samples generated by any member of a rich family of distributions, they are guaranteed to converge to the optimal separator  ...  One important domain to which the detection of low-density linear data separators is relevant is semi-supervised learning [7] .  ... 
arXiv:0805.2891v2 fatcat:55j6a3rleza6bcy2pax4pn74gu

Combining Low-Density Separators with CNNs

Yu-Xiong Wang, Martial Hebert
2016 Neural Information Processing Systems  
We propose an unsupervised margin maximization that jointly estimates compact high-density regions and infers low-density separators.  ...  By encouraging these units to learn diverse sets of low-density separators across the unlabeled data, we capture a more generic, richer description of the visual world, which decouples these units from  ...  To learn w s 's as low-density separators, we are supposed to have certain high-density regions which w s 's separate. However, accurate estimation of high-density regions is difficult.  ... 
dblp:conf/nips/WangH16 fatcat:q4dq4gbs2jgovdmy2bqc5el2aa

Semi-supervised learning objectives as log-likelihoods in a generative model of data curation [article]

Stoil Ganev, Laurence Aitchison
2021 arXiv   pre-print
We currently do not have an understanding of semi-supervised learning (SSL) objectives such as pseudo-labelling and entropy minimization as log-likelihoods, which precludes the development of e.g.  ...  CONCLUSION We showed that low-density separation SSL objectives can be understood as a lower-bound on a log-probability which arises from a principled generative model of data curation.  ...  LOW-DENSITY SEPARATION SEMI-SUPERVISED LEARNING OBJECTIVES The intuition behind low-density separation objectives for semi-supervised learning is that decision boundaries should be in low-density regions  ... 
arXiv:2008.05913v2 fatcat:xc3tqmyrpnbcpbs7hvdeiq55gm

An enhanced self-organizing incremental neural network for online unsupervised learning

Shen Furao, Tomotaka Ogura, Osamu Hasegawa
2007 Neural Networks  
Neural Networks, 19, 90-106] in the following respects: (1) it adopts a single-layer network to take the place of the two-layer network structure of SOINN; (2) it separates clusters with high-density overlap  ...  An incremental network for on-line unsupervised classification and topology learning.  ...  density of this node is low.  ... 
doi:10.1016/j.neunet.2007.07.008 pmid:17826947 fatcat:soof72sb75fofhgpqjeixepqnu

Towards Making Unlabeled Data Never Hurt

Yu-Feng Li, Zhi-Hua Zhou
2015 IEEE Transactions on Pattern Analysis and Machine Intelligence  
Unlike S3VMs which typically aim at approaching an optimal low-density separator, S4VMs try to exploit the candidate low-density separators simultaneously to reduce the risk of identifying a poor separator  ...  It is desired to have safe semisupervised learning approaches which never degenerate learning performance by using unlabeled data.  ...  Among popular semi-supervised learning approaches, S3VMs (Vapnik, 1998; Bennett & Demiriz, 1999; Joachims, 1999) are based on the low-density assumption and try to learn a low-density separator which  ... 
doi:10.1109/tpami.2014.2299812 pmid:26353217 fatcat:sbrfdjrhtfhmvbgvvzgubgblna

Deep Low-Density Separation for Semi-supervised Classification [chapter]

Michael C. Burkhart, Kyle Shan
2020 Lecture Notes in Computer Science  
In this paper, we introduce a novel hybrid method that instead applies low-density separation to the embedded features.  ...  We describe it in detail and discuss why low-density separation may be better suited for SSL on neural network-based embeddings than graph-based algorithms.  ...  and low-density separation.  ... 
doi:10.1007/978-3-030-50420-5_22 fatcat:xdvl6f2qgbfindvqqldt6aycse

A SVM Active Learning Algorithm Based on Class Boundary Characteristics, and Its Application in Audio Classification

Yan Leng, Nai Zhou, Chengli Sun, Xinyan Xu, Qi Yuan, Yunxia Liu, Dengwang Li, Zhiyuan Guo
2017 International Journal of Multimedia and Ubiquitous Engineering  
We summarize 3 characteristics of class boundary, i.e. 1) the class boundary lies in a low-density region; 2) the class boundary region is confusing; 3) there exists redundancy in the class boundary region  ...  We use the proposed active learning algorithm to resolve the sample labeling problem of audio classification.  ...  Cluster-assumption is equivalent to the low density separation assumption, and then according to the low density separation assumption, cluster-assumption can be re-expressed as: the class boundary should  ... 
doi:10.14257/ijmue.2017.12.4.06 fatcat:jiqizpmksjdcvh5sb7smb5jhey

A hierarchical self-organizing feature map for analysis of not well separable clusters of different feature density

Stefan Schünemann, Bernd Michaelis
1999 The European Symposium on Artificial Neural Networks  
A de nition of a special Mahalanobis space of the data set during the learning improves the properties concerning clusters with low density.  ...  The continuation of the learning phase for the mapk=1including adaptation towards not well separable clusters with low feature density is described in section 5.  ...  Therefore, it is possible to adapt some neurons in the direction of not well separable clusters with low feature density.  ... 
dblp:conf/esann/SchunemannM99 fatcat:z5htnjfdbjbplgkzkhrtpszkz4

A Bayesian framework for active learning

Richard Fredlund, Richard M. Everson, Jonathan E. Fieldsend
2010 The 2010 International Joint Conference on Neural Networks (IJCNN)  
We describe a Bayesian framework for active learning for non-separable data, which incorporates a query density to explicitly model how new data is to be sampled.  ...  We show the efficacy of this algorithm on the probabilistic high-low game which is a non-separable generalisation of the separable high-low game introduced by Seung et al.  ...  We consider a straightforward non-separable extension of this model, which we call the probabilistic high-low game.  ... 
doi:10.1109/ijcnn.2010.5596917 dblp:conf/ijcnn/FredlundEF10 fatcat:tcks3u5e6rhydikexmvrsvp6wi

Cross-language Neighborhood Effects in Learners Indicative of an Integrated Lexicon

Gabriela Meade, Katherine J. Midgley, Ton Dijkstra, Phillip J. Holcomb
2018 Journal of Cognitive Neuroscience  
ERPs recorded during a language decision task before and after learning also showed differences as a function of L1 neighborhood density.  ...  of learning.  ...  low-density neighborhood names (dotted).  ... 
doi:10.1162/jocn_a_01184 pmid:28880767 pmcid:PMC6088240 fatcat:ausdtsurfvgn5a6u7felhs7eji

Comparing EM Clustering Algorithm with Density Based Clustering Algorithm Using WEKA Tool

2016 International Journal of Science and Research (IJSR)  
In Density based clustering, clusters are dense regions in the data space, separated by regions of lower object density.  ...  Machine learning is type of artificial intelligence wherein computers make predictions based on data.  ...  In density-based clustering algorithms, dense areas of objects in the data space are considered as clusters, which are segregated by low-density area (noise).  ... 
doi:10.21275/v5i7.art2016420 fatcat:xrl4kt4otfgq5nbez66fb3vzjy

Learning the Number of Clusters in Self Organizing Map [chapter]

Guenael Cabanes, Younes Bennani
2010 Self-Organizing Maps  
The proposed algorithm DS2L-SOM locates regions of high density that are separated from one another by regions of low density.  ...  clusters (low density regions).  ...  /learning-the-number-of-clusters-in-self-organizingmap  ... 
doi:10.5772/9164 fatcat:sywdw3yarjc2rgzkxzv3s2uovi

SuperMeshing: A Novel Method for Boosting the Mesh Density in Numerical Computation within 2D Domain [article]

Handing Xu, Zhenguo Nie, Qingfeng Xu, Xinjun Liu
2021 arXiv   pre-print
Based on the low mesh-density stress field in the 2D plane strain problem computed by the finite element method, this method utilizes a deep neural network named SuperMeshingNet to learn the non-linear  ...  mapping from low mesh-density to high mesh-density stress field, and realizes the improvement of numerical computation accuracy and efficiency simultaneously.  ...  a pair of low mesh-density and high mesh-density stress field).  ... 
arXiv:2104.01138v1 fatcat:z2csav2fszc47it2wjq7z3nqgi

Music FaderNets: Controllable music generation based on high-Level features via low-level feature modelling

HAO HAO TAN, Dorien Herremans
2020 Zenodo  
Furthermore, we demonstrate that the model successfully learns the intrinsic relationship between arousal and its corresponding low-level attributes (rhythm and note density), with only 1% of the training  ...  We refer to our proposed framework as Music FaderNets, which is inspired by the fact that low-level attributes can be continuously manipulated by separate "sliding faders" through feature disentanglement  ...  This motivates us to use latent variable models [7] as we can learn separate latent spaces for each low-level feature to obtain disentangled controllability.  ... 
doi:10.5281/zenodo.4245375 fatcat:5i6snjwrtrhzvlnz5aegnnff4i

Music FaderNets: Controllable Music Generation Based On High-Level Features via Low-Level Feature Modelling

Hao Hao Tan, Dorien Herremans
2020 arXiv   pre-print
Furthermore, we demonstrate that the model successfully learns the intrinsic relationship between arousal and its corresponding low-level attributes (rhythm and note density), with only 1% of the training  ...  We refer to our proposed framework as Music FaderNets, which is inspired by the fact that low-level attributes can be continuously manipulated by separate "sliding faders" through feature disentanglement  ...  This motivates us to use latent variable models [7] as we can learn separate latent spaces for each low-level feature to obtain disentangled controllability.  ... 
arXiv:2007.15474v1 fatcat:tbzdhhyzmrb3vkgnfp2j346lj4
« Previous Showing results 1 — 15 out of 470,791 results