One-Sided Unsupervised Domain Mapping
Authors: Sagie Benaim, Lior Wolf
NeurIPS 2017 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We present experimental results that the new method not only allows for one sided mapping learning, but also leads to preferable numerical results over the existing circularity-based constraint. 4 Experiments We compare multiple methods: the Disco GAN or the Cycle GAN baselines; the one sided mapping using Ldistance (A B or B A); the combination of the baseline method with Ldistance; the self distance method. |
| Researcher Affiliation | Collaboration | Sagie Benaim1 and Lior Wolf1,2 1The Blavatnik School of Computer Science , Tel Aviv University, Israel 2Facebook AI Research |
| Pseudocode | No | The paper describes the methods and loss functions in prose and mathematical notation, but does not include structured pseudocode or algorithm blocks. |
| Open Source Code | Yes | Our entire code is made publicly available at https://github.com/sagiebenaim/Distance GAN. |
| Open Datasets | Yes | The 3D car dataset [4] consists of rendered images of 3D cars... Similarly, the head dataset, [17], consists of 3D images of rotated heads... Celeb A [26, 14] was annotated for multiple attributes... We use the baseline Cycle GAN method as well as our methods in order to translate from Street View House Numbers (SVHN) [15] to MNIST [12]. |
| Dataset Splits | No | For the car2car experiment, the car dataset is split into two parts, one of which is used for A and one for B (It is further split into train and test set). This mentions train and test, but lacks specific percentages, counts, or a clear method for reproduction of the split. No explicit mention of a 'validation' set or its split details. |
| Hardware Specification | No | The paper does not provide specific details regarding the hardware used for running the experiments (e.g., GPU/CPU models, memory). |
| Software Dependencies | No | The paper mentions using architectures from Disco GAN and Cycle GAN, but does not provide specific version numbers for software dependencies or frameworks (e.g., TensorFlow, PyTorch, Python versions). |
| Experiment Setup | Yes | Table 1: Tradeoff weights for each experiment. For Disco GAN, we use a fixed weight configuration for all experiments, as shown in Tab. 1. |