Comparison of performance of autoencoder with PCA

Question

I am running PCA and autoencoder (2 hidden layer with relu) on a data. Both PCA and autoencoder give similar accuracy of the order 50%. I have tried different variations of autoencoder: changing layers, neurons,activations. But it did not perform any better. What can be the possible reason behind this. Can anybody suggest any analysis of data to identify the reason for autoencoder to not perform any better.

Answer 1

The features in your data are probably linear as an Autoencoder should give the same results then (see here and here). The surplus of an AE is that it is efficient to detect nonlinear features which a PCA cannot.