📄 contents.m
字号:
% Data Description Toolbox% Version 1.11 23-Nov-2003%%Dataset construction%--------------------%isocset true if dataset is one-class dataset%gendatoc generate a one-class dataset from two data matrices%oc_set change normal classif. problem to one-class problem%target_class extracts the target class from an one-class dataset%make_outliers create outlier data in a box around target class%gendatgrid create a grid dataset around a 2D dataset%gendatout create outlier data in a hypersphere around the% target data%dd_crossval cross-validation dataset creation%%Data preprocessing%------------------%myproxm replacement for proxm.m%kwhiten rescale data to unit variance in kernel space%gower compute the gower similarities%%One-class classifiers%---------------------%random_dd description which randomly assigns labels%gauss_dd data description using normal density%rob_gauss_dd robustified gaussian distribution%mcd_gauss_dd Minimum Covariance Determinant gaussian%mog_dd mixture of Gaussians data description%parzen_dd Parzen density data description%%autoenc_dd auto-encoder neural network data description%kcenter_dd k-center data description%kmeans_dd k-means data description%pca_dd principal component data description%som_dd Self-Organizing Map data description%%nndd nearest neighbor based data description%knndd K-nearest neighbor data description%svdd Support vector data description%ksvdd SVDD on general kernel matrices%lpdd linear programming data description%dlpdd distance-linear programming data description%%isocc true if classifier is one-class classifier%consistent_occ optimize the hyperparameter using consistency%%%Error computation.%-----------------%dd_error false positive and negative fraction of classifier%dd_f1 F1 score computation%dd_roc computation of the Receiver-Operating Characterisic curve %dd_auc error under the ROC curve%dd_delta_aic AIC error for density estimators%dd_fp compute false positives for given false negative% fraction%simple_roc basic ROC curve computation%%Plot functions.%--------------%plotroc plot an ROC curve%plotg plot a 2D grid of function values%plotw plot a 2D real-valued output of classifier w%%Support functions.%-----------------%find_target gives the indices of target and outlier objs from a dataset%getoclab returns numeric labels (+1/-1)%dist2dens map distance to posterior probabilities%dd_threshold give percentiles for a sample%setthres set the threshold for a one-class classifier%randsph create outlier data uniformly in a unit hypersphere%makegriddat auxiliary function for constructing grid data%relabel relabel a dataset%center center the kernel matrix in kernel space%gausspdf multi-variate Gaussian prob.dens.function%mahaldist Mahalanobis distance%sqeucldistm square Euclidean distance%mogEM EM procedure to optimize Mixture of Gaussians%mogP probability density of Mixture of Gaussians%mykmeans own implementation of the k-means clustering algorithm%getfeattype find the nominal and continuous features%knn_optk optimization of k for the knndd using leave-one-out%volsphere compute the volume of a hypersphere%scale_range compute a reasonable range of scales for a dataset%%%Examples%--------%dd_ex1 show performance of nndd and svdd%dd_ex2 show the performances of a list of classifiers%dd_ex3 shows the use of the svdd and ksvdd%dd_ex4 optimizes a hyperparameter using consistent_occ%dd_ex5 shows the construction of lpdd from dlpdd%% Copyright: D.M.J. Tax, davidt@ph.tn.tudelft.nl% Faculty of Applied Physics, Delft University of Technology% P.O. Box 5046, 2600 GA Delft, The Netherlands
⌨️ 快捷键说明
复制代码
Ctrl + C
搜索代码
Ctrl + F
全屏模式
F11
切换主题
Ctrl + Shift + D
显示快捷键
?
增大字号
Ctrl + =
减小字号
Ctrl + -