Classification Asymptotics in the Random Matrix Regime. Couillet, R., Liao, Z., & Mai, X. In 2018 26th European Signal Processing Conference (EUSIPCO), pages 1875-1879, Sep., 2018. Paper doi abstract bibtex This article discusses the asymptotic performance of classical machine learning classification methods (from discriminant analysis to neural networks) for simultaneously large and numerous Gaussian mixture modelled data. We first provide theoretical bounds on the minimally discriminable class means and covariances under an oracle setting, which are then compared to recent theoretical findings on the performance of machine learning. Non-obvious phenomena are discussed, among which surprising phase transitions in the optimal performance rates for specific hyperparameter settings.
@InProceedings{8553034,
author = {R. Couillet and Z. Liao and X. Mai},
booktitle = {2018 26th European Signal Processing Conference (EUSIPCO)},
title = {Classification Asymptotics in the Random Matrix Regime},
year = {2018},
pages = {1875-1879},
abstract = {This article discusses the asymptotic performance of classical machine learning classification methods (from discriminant analysis to neural networks) for simultaneously large and numerous Gaussian mixture modelled data. We first provide theoretical bounds on the minimally discriminable class means and covariances under an oracle setting, which are then compared to recent theoretical findings on the performance of machine learning. Non-obvious phenomena are discussed, among which surprising phase transitions in the optimal performance rates for specific hyperparameter settings.},
keywords = {Gaussian processes;learning (artificial intelligence);mixture models;pattern classification;classification asymptotics;random matrix regime;classification methods;discriminant analysis;neural networks;minimally discriminable class means;covariances;oracle setting;machine learning classification methods;Gaussian mixture modelled data;Machine learning;Covariance matrices;Kernel;Support vector machines;Europe;Signal processing;Neural networks;Random matrix theory;classification;kernel methods;neural networks;LDA/QDA},
doi = {10.23919/EUSIPCO.2018.8553034},
issn = {2076-1465},
month = {Sep.},
url = {https://www.eurasip.org/proceedings/eusipco/eusipco2018/papers/1570435395.pdf},
}
Downloads: 0
{"_id":"QESsv7pycEWaeM3ix","bibbaseid":"couillet-liao-mai-classificationasymptoticsintherandommatrixregime-2018","authorIDs":[],"author_short":["Couillet, R.","Liao, Z.","Mai, X."],"bibdata":{"bibtype":"inproceedings","type":"inproceedings","author":[{"firstnames":["R."],"propositions":[],"lastnames":["Couillet"],"suffixes":[]},{"firstnames":["Z."],"propositions":[],"lastnames":["Liao"],"suffixes":[]},{"firstnames":["X."],"propositions":[],"lastnames":["Mai"],"suffixes":[]}],"booktitle":"2018 26th European Signal Processing Conference (EUSIPCO)","title":"Classification Asymptotics in the Random Matrix Regime","year":"2018","pages":"1875-1879","abstract":"This article discusses the asymptotic performance of classical machine learning classification methods (from discriminant analysis to neural networks) for simultaneously large and numerous Gaussian mixture modelled data. We first provide theoretical bounds on the minimally discriminable class means and covariances under an oracle setting, which are then compared to recent theoretical findings on the performance of machine learning. Non-obvious phenomena are discussed, among which surprising phase transitions in the optimal performance rates for specific hyperparameter settings.","keywords":"Gaussian processes;learning (artificial intelligence);mixture models;pattern classification;classification asymptotics;random matrix regime;classification methods;discriminant analysis;neural networks;minimally discriminable class means;covariances;oracle setting;machine learning classification methods;Gaussian mixture modelled data;Machine learning;Covariance matrices;Kernel;Support vector machines;Europe;Signal processing;Neural networks;Random matrix theory;classification;kernel methods;neural networks;LDA/QDA","doi":"10.23919/EUSIPCO.2018.8553034","issn":"2076-1465","month":"Sep.","url":"https://www.eurasip.org/proceedings/eusipco/eusipco2018/papers/1570435395.pdf","bibtex":"@InProceedings{8553034,\n author = {R. Couillet and Z. Liao and X. Mai},\n booktitle = {2018 26th European Signal Processing Conference (EUSIPCO)},\n title = {Classification Asymptotics in the Random Matrix Regime},\n year = {2018},\n pages = {1875-1879},\n abstract = {This article discusses the asymptotic performance of classical machine learning classification methods (from discriminant analysis to neural networks) for simultaneously large and numerous Gaussian mixture modelled data. We first provide theoretical bounds on the minimally discriminable class means and covariances under an oracle setting, which are then compared to recent theoretical findings on the performance of machine learning. Non-obvious phenomena are discussed, among which surprising phase transitions in the optimal performance rates for specific hyperparameter settings.},\n keywords = {Gaussian processes;learning (artificial intelligence);mixture models;pattern classification;classification asymptotics;random matrix regime;classification methods;discriminant analysis;neural networks;minimally discriminable class means;covariances;oracle setting;machine learning classification methods;Gaussian mixture modelled data;Machine learning;Covariance matrices;Kernel;Support vector machines;Europe;Signal processing;Neural networks;Random matrix theory;classification;kernel methods;neural networks;LDA/QDA},\n doi = {10.23919/EUSIPCO.2018.8553034},\n issn = {2076-1465},\n month = {Sep.},\n url = {https://www.eurasip.org/proceedings/eusipco/eusipco2018/papers/1570435395.pdf},\n}\n\n","author_short":["Couillet, R.","Liao, Z.","Mai, X."],"key":"8553034","id":"8553034","bibbaseid":"couillet-liao-mai-classificationasymptoticsintherandommatrixregime-2018","role":"author","urls":{"Paper":"https://www.eurasip.org/proceedings/eusipco/eusipco2018/papers/1570435395.pdf"},"keyword":["Gaussian processes;learning (artificial intelligence);mixture models;pattern classification;classification asymptotics;random matrix regime;classification methods;discriminant analysis;neural networks;minimally discriminable class means;covariances;oracle setting;machine learning classification methods;Gaussian mixture modelled data;Machine learning;Covariance matrices;Kernel;Support vector machines;Europe;Signal processing;Neural networks;Random matrix theory;classification;kernel methods;neural networks;LDA/QDA"],"metadata":{"authorlinks":{}}},"bibtype":"inproceedings","biburl":"https://raw.githubusercontent.com/Roznn/EUSIPCO/main/eusipco2018url.bib","creationDate":"2021-02-13T15:38:40.080Z","downloads":0,"keywords":["gaussian processes;learning (artificial intelligence);mixture models;pattern classification;classification asymptotics;random matrix regime;classification methods;discriminant analysis;neural networks;minimally discriminable class means;covariances;oracle setting;machine learning classification methods;gaussian mixture modelled data;machine learning;covariance matrices;kernel;support vector machines;europe;signal processing;neural networks;random matrix theory;classification;kernel methods;neural networks;lda/qda"],"search_terms":["classification","asymptotics","random","matrix","regime","couillet","liao","mai"],"title":"Classification Asymptotics in the Random Matrix Regime","year":2018,"dataSources":["yiZioZximP7hphDpY","iuBeKSmaES2fHcEE9"]}