non-linear dimension reduction via Autoencoder

Asked by abassidodo on 5 Oct 2017

Latest activity Answered by BERGHOUT Tarek on 11 Apr 2019

13 views (last 30 days)

hello all, I am trying to use the Matlab implementation of autoencoder to reduce the dimension of 1509 samples of Bag-of-visual word models of images, but I am surprised that while the image classification without dimension reduction recorded about 50% accuracy, and Matlab's PCA improved it to 60% but the Matlab implementation of autoencoder (with logsig activation and default values for all the parameters) reduced it to 40%. I expect higher accuracy from autoencoder, what can be the problem?

1 Comment

John D'Errico 9 Mar 2019

@preksha - Actually, no, they cannot share it. You cannot give a function from a toolbox to another person to then use it if they lack a license. And if you have a license, then you don't need it.

Since it seems that the autoencoder class was introduced with R2015b, all you then need is the appropriate toolbox (looks like the Deep Learning toolbox.)

1 Answer

Link

BERGHOUT Tarek 님의 답변 11 Apr 2019

1) try to normalize you data first, between 0 and 1.

2) use these autoencoders and tell me the difference

https://www.mathworks.com/matlabcentral/fileexchange/66080-auto-encoder-based-extreme-learning-machine?s_tid=prof_contriblnk

You are now following this question

abassidodo (view profile)