Autoencoding beyond pixels using a learned similarity metric
Authors: Anders Boesen Lindbo Larsen, Søren Kaae Sønderby, Hugo Larochelle, Ole Winther
ICML 2016 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | In this section we investigate the performance of different generative models: Plain VAE with an element-wise Gaussian observation model. VAE with a learned distance (VAEDisl). The combined VAE/GAN model. A GAN. We apply our methods to face images from the Celeb A dataset2 (Liu et al., 2015). |
| Researcher Affiliation | Collaboration | Anders Boesen Lindbo Larsen1 ABLL@DTU.DK Søren Kaae Sønderby2 SKAAESONDERBY@GMAIL.COM Hugo Larochelle3 HLAROCHELLE@TWITTER.COM Ole Winther1,2 OLWI@DTU.DK 1 Department for Applied Mathematics and Computer Science, Technical University of Denmark 2 Bioinformatics Centre, Department of Biology, University of Copenhagen, Denmark 3 Twitter, Cambridge, MA, USA |
| Pseudocode | Yes | Algorithm 1 Training the VAE/GAN model |
| Open Source Code | Yes | We refer to our implementation available online1. 1http://github.com/andersbll/ autoencoding_beyond_pixels |
| Open Datasets | Yes | We apply our methods to face images from the Celeb A dataset2 (Liu et al., 2015). |
| Dataset Splits | No | The paper mentions a 'training set' and 'test set' for the LFW dataset, but does not explicitly provide details about a 'validation' split or specify percentages/counts for any data splits. |
| Hardware Specification | No | The paper mentions 'Nvidia for donating GPUs used in experiments' but does not specify the exact GPU model or any other hardware components like CPU, memory, or specific machine configurations. |
| Software Dependencies | No | The paper mentions using 'Deep Py3 and CUDArray' as software frameworks and cites a technical report for CUDArray, but it does not specify version numbers for these or any other software dependencies. |
| Experiment Setup | Yes | Our models are trained with RMSProp using a learning rate of 0.0003 and a batch size of 64. In table 1 we list the network architectures. |