/autoencoder

A Stacked Autoencoder that is Trained via Parallelized Evolutionary Optimization

Primary LanguageTeXMIT LicenseMIT

Stacked Denoising Autoencoders by Jason Liang and Keith Kelly

Features: 1)C++ implementation of feed-forward neural networks and stacked denoising autoencoders 2)MultiThreaded and Fast: Order of magnitude faster than training with Theano

To build with OpenMP: 1)cd to src directory and type "make"

To build with OpenBlas (you must have it installed): 1)cd to src directory 2)in the Makefile, set OPEN_BLAS_INC and OPEN_BLAS_LIB appropriately 3)type "make HAS_OPENBLAS=1"

Project Report: http://users.ices.utexas.edu/~keith/files/autoencoder/final_report/autoencoder.pdf

Datasets can be found at: users.ices.utexas.edu/~keith/files/MLProject.html