To use the function "TRLDA" or "ODLDA", please follow the input/output format:
[ W ] = TRLDA( Sw,Sb,m,alpha,d,t) [ WW,W1,s,fw ] = ODLDA( Sw,Sb,d,t)
Sw ： dd intra-class scatter matrix Sb : dd inter-class scatter matrix, d is the dimension of samples m : subspace dimension, m < d alpha : Relaxation factor t : iteration times W ： dm projection matrix WW : 1d cell array, i-th cell indicates the projection matrix W ∈ R^{di} W1 : projection matrix with the optimal subspace dimension s ： ((Tr(W'SbW))^0.5)/(Tr(W'SwW)) fw ： dt, the objective function values
Please make sure that the documents Eu2_distance.m and ClusteringMeasure.m are in the same folder as TRLDA.m and ODLDA.m
Use the codes, please cite Wang J, Wang L, Nie F, et al. A novel formulation of trace ratio linear discriminant analysis[J]. IEEE Transactions on Neural Networks and Learning Systems, 33(10), pp. 5568-5578, 2022.
If you have any questions, please connect wanglinjun@mail.nwpu.edu.cn