Yet ANother autoencoder toolbox based on Theano
Switch branches/tags
Nothing to show
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Failed to load latest commit information.


Yet ANother autoencoder toolbox based on Theano

We implemented the autoencoder (AE) algorithm in Python based on the Theano library. The materials in helped a lot. the shallow AE. Various strategies can be enabled or disabled, including denoising, sparsity, L1/L2 regularization, and weight tying strategies. One can also config the loss function, activation function, and training algorithm (e.g. minibatch, CG). the stacked AE (SAE), built upon the shallow ones. Methods include layer-wise pretraining, unsupervised and supervised fine-tuning. multi-layer perceptron (fully connected network), mainly implemented for fine-tuning of SAE. marginalized denoising autoencoder. drift correction autoencoder (DCAE), designed for correction of instrumental variation and time-varying drift in sensor systems. Please see the ref for details. test SAE and MSDA on benchmark datasets.

** test DCAE on the corn dataset. test DCAE on the gas sensor array drift dataset. Description of these two datasets is in datasets_readme.txt.

.idea: project files of PyCharm.

ref: Ke Yan, and David Zhang, "Correcting Instrumental Variation and Time-varying Drift: A Transfer Learning Approach with Autoencoders, " accepted by Instrumentation and Measurement, IEEE Transactions on

Copyright 2016 YAN Ke, Tsinghua Univ. ,