Auxiliary Deep Generative Models

Authors: Lars Maaløe, Casper Kaae Sønderby, Søren Kaae Sønderby, Ole Winther

ICML 2016 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We show state-of-the-art performance within semi-supervised learning on MNIST, SVHN and NORB datasets.
Researcher Affiliation Academia 1Department of Applied Mathematics and Computer Science, Technical University of Denmark 2Bioinformatics Centre, Department of Biology, University of Copenhagen
Pseudocode No The paper describes models and training procedures mathematically and textually, but it does not include any explicit pseudocode or algorithm blocks.
Open Source Code Yes Implementation is available in a repository named auxiliary-deep-generative-models on github.com.
Open Datasets Yes For the benchmark datasets we show (cf. Sec. 4.4): (iii) State-of-the-art results on several semi-supervised classification tasks. We demonstrate the benefits of the increased flexibility by achieving state-of-the-art performance in the semi-supervised setting for the MNIST (Le Cun et al., 1998), SVHN (Netzer et al., 2011) and NORB (Le Cun et al., 2004) datasets.
Dataset Splits Yes For the MNIST dataset we have combined the training set of 50000 examples with the validation set of 10000 examples. The test set remained as is.
Hardware Specification Yes This research was supported by the Novo Nordisk Foundation, Danish Innovation Foundation and the NVIDIA Corporation with the donation of TITAN X and Tesla K40 GPUs.
Software Dependencies No The models are implemented in Python using Theano (Bastien et al., 2012), Lasagne (Dieleman et al., 2015) and Parmesan libraries
Experiment Setup Yes For training, we have used the Adam (Kingma and Ba, 2014) optimization framework with a learning rate of 3e4, exponential decay rate for the 1st and 2nd moment at 0.9 and 0.999, respectively. The β constant was between 0.1 and 2 throughout the experiments.