A Forest from the Trees: Generation through Neighborhoods
Authors: Yang Li, Tianxiang Gao, Junier Oliva4755-4762
AAAI 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | The efficacy of our model is shown empirically with standard image datasets. We observe compelling results and a significant improvement over baselines. Combined further with a contrastive training mechanism, our proposed methods can effectively perform non-parametric novelty detection. 3 Experiments |
| Researcher Affiliation | Academia | Yang Li, Tianxiang Gao, Junier B. Oliva Department of Computer Science, UNC Chapel Hill {yangli95, tianxiang, joliva}@cs.unc.edu |
| Pseudocode | No | The paper describes the methods using equations and textual explanations, but it does not include structured pseudocode or algorithm blocks. |
| Open Source Code | No | The paper does not provide an explicit statement about releasing source code or a direct link to a code repository for the methodology described. It only provides a link to an arXiv preprint ("https://arxiv.org/abs/1902.01435") for a full version of the paper. |
| Open Datasets | Yes | We show the efficacy of our models on standard benchmark image datasets. We use the original training-testing split of MNIST dataset. Training is conducted using data of a single class from original training split. The paper references specific datasets like MNIST, SVHN, CIFAR-10, and Celeb A, which are widely recognized public datasets. |
| Dataset Splits | No | The paper states "We use the original training-testing split of MNIST dataset" and mentions "early stopping to prevent overfitting", which implies the use of a validation set, but it does not specify the exact split percentages, sample counts, or methodology for the validation set itself. |
| Hardware Specification | No | The paper does not provide specific hardware details such as exact GPU/CPU models, processor types, or memory amounts used for running the experiments. It does not mention any cloud or cluster resources with specifications. |
| Software Dependencies | No | The paper mentions using "Real NVP" and "Pixel CNN" as models and baselines but does not specify any ancillary software dependencies (e.g., programming languages, libraries, frameworks) with version numbers that would be needed to replicate the experiments. |
| Experiment Setup | Yes | Following the exact preprocessing procedure in Real NVP, we transform the pixels into logit space to alleviate the impact of boundary effects. In order to conduct a fair comparison with Real NVP, we use exactly the same network architecture and hyperparameters as those in (Dinh, Sohl-Dickstein, and Bengio 2016). We do not apply data augmentation for all experiments, but we use early stopping to prevent overfitting. The margin in contrastive loss is set to 0.5 bits per dimension. To train our neighbor based models, we pull neighbors by Euclidean distance on PCA features, 5 neighbors are used here. |