Eliminating accidental deviations to minimize generalization error and maximize replicability: Applications in connectomics and genomics.
Eric W. BridgefordShangsi WangZeyi WangTing XuR. Cameron CraddockJayanta DeyGregory KiarWilliam Gray-RoncalCarlo ColantuoniChristopher DouvilleStephanie NobleCarey E. PriebeBrian CaffoMichael MilhamXi-Nian ZuoJoshua T. VogelsteinPublished in: PLoS Comput. Biol. (2021)
Keyphrases
- generalization error
- cross validation
- model selection
- training data
- learning algorithm
- binary classification
- active learning
- upper bound
- linear classifiers
- classification error
- training set
- sample complexity
- sample size
- supervised learning
- expected error
- target function
- training error
- conditional expectation
- uniform convergence
- generalization error bounds
- perceptron learning
- boosting algorithms
- prediction accuracy
- learning machines
- training set size
- text mining
- subspace information criterion
- support vector
- multi class