search

lijin118 / lctc

Heterogeneous domain adaptation through progressive alignment, TNNLS
17 stars 2 forks source link

Cannot reproduce the performance of SVMt in the paper #3

Open hellowangqian opened 4 years ago

hellowangqian commented 4 years ago

I'm wondering if the authors would like to release the code and data for the baseline method SVMt. I've tried using libsvm with linear kernel and other default parameters to reproduce the results in the paper but only got results as follows (mean and std over 10 random trials): Caltech: 24.06 (6.4); Amazon: 37.26 (5.23); W: 54.72 (5.68). These results are worse than those reported in the paper (Caltech: 30.1 (0.9); Amazon: 44.2(1.1); W: 58.3(1.2)) in terms of both mean and std. I'd like to know what leads to such difference. Any comments will be much appreciated.

LilyQahh0208 commented 4 years ago

Hello, when I run the code, an error occurs. "Incorrect use of pca, The value of X must not be sparse". Could you please tell me how to solve it? Thank you very much.

hellowangqian commented 4 years ago

Hello, when I run the code, an error occurs. "Incorrect use of pca, The value of X must not be sparse". Could you please tell me how to solve it? Thank you very much.

As said in the message, X should not be sparse: X = full(X) may solve this.

LilyQahh0208 commented 4 years ago

Thanks for your reply and I have solved this problem. However, There is another error. "未定义与 'double' 类型的输入参数相对应的函数 'train'"(model = train(Yref,sparse(St(:,1:size(Xref,1))'),sprintf('-s 0 -c %f -q 1',100));). The online solution is because the train function file is not in the current path, so I copied the train function file into the current path, but another error occured. "Incorrect use of train.First argument is not a neural network." I am not good at matlab.Could you help me how to solve it? Thanks a lot.

hellowangqian commented 4 years ago

The error is caused by missing function “train” which is for training a SVM. You should have it in this folder: liblinear/matlab And make sure add this folder to the path using addpath or manually from the matlab menu.

Also check out our latest code for HDA here: https://github.com/hellowangqian/cdspp-hda

Sent from Mailhttps://go.microsoft.com/fwlink/?LinkId=550986 for Windows 10

From: LilyQahh0208mailto:notifications@github.com Sent: 03 August 2020 10:39 To: lijin118/lctcmailto:lctc@noreply.github.com Cc: hellowangqianmailto:wangqian.sunny@hotmail.com; Authormailto:author@noreply.github.com Subject: Re: [lijin118/lctc] Cannot reproduce the performance of SVMt in the paper (#3)

Thanks for your reply and I have solved this problem. However, There is another error. "未定义与 'double' 类型的输入参数相对应的函数 'train'"(model = train(Yref,sparse(St(:,1:size(Xref,1))'),sprintf('-s 0 -c %f -q 1',100));). The online solution is because the train function file is not in the current path, so I copied the train function file into the current path, but another error occured. "Incorrect use of train.First argument is not a neural network." I am not good at matlab.Could you help me how to solve it? Thanks a lot.

― You are receiving this because you authored the thread. Reply to this email directly, view it on GitHubhttps://github.com/lijin118/lctc/issues/3#issuecomment-667920909, or unsubscribehttps://github.com/notifications/unsubscribe-auth/ABX467TDJ7O7UDHKLGOSGSLR62AWNANCNFSM4J7FEYZQ.

LilyQahh0208 commented 4 years ago

Thank you so much for your patience to answer,I have solved the problem, the code can run successfully. By the way,do these two lines of code "addpath(genpath('../ml'));addpath(genpath('./tSNE_matlab'));" contain the necessary packages? I got the "constructW.m" file and "EuDist2.m" file from the web to run the code.

Morrie0804 commented 3 years ago

@LilyQahh0208 Could you please tell me where I can find "constructW.m" file and "EuDist2.m" file? Thanks a lot.